- Email: [email protected]
Effectiveness and cost-effectiveness of implantable cardioverter defibrillators in the treatment of ventricular arrhythmias among medicare beneficiaries
CLINICAL STUDIES
Effectiveness and Cost-effectiveness of Implantable Cardioverter Defibrillators in the Treatment of Ventricular Arrhythmias among Medicare Beneficiaries J. Peter Weiss, MD, Olga Saynina, MA, MBA, Kathryn M. McDonald, MM, Mark B. McClellan, MD, PhD, Mark A. Hlatky, MD PURPOSE: The implantable cardioverter defibrillator has been assessed in randomized trials, but the generalizability of trial results to broader clinical settings is unclear. Our purpose was to evaluate the outcomes and costs of defibrillator use in an unselected population. SUBJECTS AND METHODS: We identified 125,892 Medicare patients who were discharged between 1987 and 1995 after hospitalization with a primary diagnosis of ventricular tachycardia or ventricular fibrillation, 7789 of whom (6.2%) received a defibrillator. We used a multivariable propensity score that included patient and hospital characteristics to match pairs of patients, in which one patient received a defibrillator and the other did not. We compared mortality and costs in these 7612 matched pairs during 8 years of follow-up. RESULTS: Patients who received a defibrillator were more likely to be younger, white, male, and urban dwelling, and to have ischemic heart disease, heart failure, or a history of ven-
tricular fibrillation. In the matched-pairs analysis, those who received a defibrillator had significantly lower mortality: 11% versus 19% at 1 year (odds ratio [OR] ⫽ 0.57; 95% confidence interval [CI]: 0.51 to 0.63), 20% versus 30% at 2 years (OR ⫽ 0.66; 95% CI: 0.60 to 0.72), and 28% versus 39% at 3 years (OR ⫽ 0.70; 95% CI: 0.63 to 0.77). These patients also had lower mortality at 8 years (P ⫽ 0.0001), although this advantage over patients who received medical treatment only decreased over time. Expenditures among defibrillator recipients were consistently higher, with a cost-effectiveness ratio of $78,400 per lifeyear gained. CONCLUSION: The use of implantable defibrillators was associated with significantly lower mortality and higher costs, whereas the cost-effectiveness was higher than many, but not all, generally accepted therapies. Am J Med. 2002;112:519 –527. 䉷2002 by Excerpta Medica, Inc.
C
26,000 implantations performed in the United States in 1998 (1). The effectiveness of the implantable defibrillator in reducing mortality has been shown in randomized clinical trials (4 – 6). However, the patients in these trials were carefully selected and were treated in medical centers that had facilities for performing specific cardiac procedures, and the results therefore may not be generalized to less selected patients in more typical practice settings. In view of these limitations, the expense associated with implantation and maintenance of the device, and the number of patients who could benefit from its use, we sought to assess the clinical outcomes and cost-effectiveness of the implantable defibrillator in a population-based cohort of Medicare patients.
ardiac arrest affects an estimated 1000 persons each day in the United States and leads to 220,000 deaths each year (1). Sudden death accounts for nearly a third of all cardiac deaths and for up to half of deaths related to ischemic heart disease (2). Patients who have had out-of-hospital cardiac arrest are particularly at high risk (3), but their outcomes have improved considerably in recent years, particularly with the development of the implantable cardioverter defibrillator. The implantable defibrillator is widely used, with approximately
From the Department of Health Research and Policy (JPW, MAH), Stanford, California; and Department of Medicine (KMM, MBM), Stanford University School of Medicine, Stanford, California; and the National Bureau of Economics Research (OS, MBM), Cambridge, Massachusetts. Supported by Grant HS 08362 from the Agency for Healthcare Research and Quality, Rockville, Maryland. Requests for reprints should be addressed to Mark A. Hlatky, MD, Stanford University School of Medicine, HRP Redwood Building, Room 150, Stanford, California 94305-5405. Manuscript submitted June 11, 2001, and accepted in revised form January 18, 2002. 䉷2002 by Excerpta Medica, Inc. All rights reserved.
MATERIALS AND METHODS We used the Health Care Financing Administration Medicare Provider Analysis and Review inpatient hospitalization file to identify all patients who were discharged 0002-9343/02/$–see front matter 519 PII S0002-9343(02)01078-1
Effectiveness and Cost-effectiveness of Implantable Defibrillators/Weiss et al
from a U.S. hospital between January 1, 1987, and September 30, 1995, and who had a primary discharge diagnosis of ventricular fibrillation or ventricular tachycardia (International Classification of Diseases, Ninth Revision, Clinical Modification [ICD-9-CM] codes 427.1, 427.4, 427.5). Patients who had been hospitalized for ventricular fibrillation or ventricular tachycardia between January 1, 1984, and December 31, 1986, were excluded. We began inclusion in 1987 because it was the first year in which at least 1% of the patients in the cohort received an implantable defibrillator. Recipients were defined as patients who received a device within 90 days of hospital admission (ICD-9-CM procedure codes 37.94 to 37.96); patients who underwent generator replacements or lead revisions were excluded. To ensure the stability of the cohort (7–9), we used data from the Medicare Beneficiary Health Insurance Skeletonized Eligibility Write-off file to exclude patients who were younger than 65 years; did not reside in the United States; were enrolled in a health maintenance organization; had noncontinuous enrollment in Medicare Part A or Part B during the 1 year following initial admission; were admitted to a non–acute care hospital, federal hospital, or hospital outside the United States; had invalid mortality data; or had missing Write-off file data. We excluded patients with possible miscoding of ventricular arrhythmia as the primary diagnosis: discharge with a total stay ⬍2 days, admittance on the same day as or the day after discharge from an admission unrelated to ventricular arrhythmia, acute myocardial infarction during index admission, or elective admission. Elective defibrillator implantations that were thought to be unrelated to the index ventricular arrhythmia event were eliminated by excluding patients who received an implant on the day of index admission or the day after admission. The cohort was limited to those discharged alive from the index admission to avoid bias due to the selection of therapy in moribund patients. These exclusions were not mutually exclusive and resulted in a final cohort of 125,892 patients. A single longitudinal record for each patient was created by linking the index admission to all previous and subsequent admissions. Expenditures were defined as the sum of diagnostic related group payments plus outlier payments, normalized to 1999 U.S. dollars using the Gross Domestic Product Deflator inflation index (http:// www.jsc.nasa.gov/bu2/inflateGDP.html).
Statistical Analysis Unadjusted analysis. All-cause mortality at 1, 2, and 3 years was compared between patients who did and did not receive a defibrillator, with results reported as odds ratios (ORs) and 95% confidence intervals (CIs). Propensity score. To address the possibility that the primary outcomes might be biased by the selection of pa520
May 2002
THE AMERICAN JOURNAL OF MEDICINE威
Volume 112
tients for treatment with an implantable defibrillator, we performed a propensity score adjustment (10,11). Demographic and clinical data thought to be predictive of treatment selection were entered as independent variables in a multivariable logistic regression that had receipt of an implantable defibrillator as the dependent variable. All variables were retained in the model, regardless of the level of statistical significance. Demographic variables included sex, race, age (in 5-year intervals), and rural versus urban residence. Comorbid conditions and secondary diagnoses included admission with a primary diagnosis of ventricular fibrillation versus ventricular tachycardia, ischemic heart disease, heart failure, prior myocardial infarction, atrial fibrillation or flutter, cardiac conduction abnormality, valvular disease, cardiomyopathy, chronic obstructive pulmonary disease, fluid or electrolyte imbalance, hypertension, diabetes, syncope, peripheral vascular disease, respiratory failure or arrest, stroke, and other central nervous system disorders. Also included were year of entry, hospital admission within the previous year for heart failure or myocardial infarction, hospital admission within the previous year for any reason, and prior coronary bypass surgery or angioplasty. We adjusted for hospital characteristics such as capability for cardiac catheterization, cardiac surgery, electrophysiologic study, and defibrillator implantation. A propensity score was calculated for each patient that yielded the probability (0 to 1) of receipt of an implantable defibrillator, based on demographic, diagnostic, and hospital characteristics. Matched-pairs analysis. Patients who received a defibrillator were matched one to one with patients who did not receive a defibrillator but who had an equivalent propensity score (12), defined as an absolute value (propensity score for defibrillator patient minus propensity score for nondefibrillator patient) ⱕ0.01. Those who did not receive a defibrillator were accepted as matches only if their survival time was equal to or greater than the time to device implantation for the corresponding defibrillator patient. Mortality for the matched cohort at 1, 2, and 3 years was estimated using logistic regression, as well as 8-year Kaplan-Meier cumulative survival. Cumulative expenditures were calculated in 1999 constant dollars, discounted at 3% per year of follow-up (13). The cumulative expenditures and mean cumulative survival (discounted at 3% per year) were used to calculate marginal cost-effectiveness as (Costdefibrillator minus Costnondefibrillator) / (LifeYearsdefibrillator minus Life-Yearsnondefibrillator). One-, 2-, and 3-year mortality for a subset of the matched pairs within the middle tertile of propensity score was also examined using logistic regression to assess the outcomes of patients with an intermediate likelihood of receiving either course of therapy. Analysis was performed using SAS software, version 6.12 (SAS Institute, Cary, North Carolina).
Effectiveness and Cost-effectiveness of Implantable Defibrillators/Weiss et al
Table 1. Characteristics of the Entire Medicare Cohort
Variable
Patients Who Received Patients Who Received a Defibrillator Medical Treatment (n ⫽ 7789) (n ⫽ 118 103) Mean ⫾ SD or Percentage
Demographic characteristics Age (years) White race Female sex Urban residence Cardiac comorbid conditions Primary diagnosis of ventricular fibrillation Ischemic heart disease Heart failure Hypertension Atrial fibrillation or flutter Valvular disease Cardiac conduction disorder Cardiomyopathy Acute myocardial infarction admission in prior year Heart failure admission in prior year Prior myocardial infarction Prior coronary angioplasty Prior bypass surgery Other comorbid conditions Fluid or electrolyte imbalance Pulmonary disease Central nervous system disorders Diabetes Syncope Respiratory failure or arrest Stroke Peripheral vascular disease Any hospitalization in prior year Hospital characteristics Angiography capability only Bypass surgery capability Electrophysiologic study/implantable defibrillator capability Procedures performed within 90 days Angiography Bypass surgery Coronary angioplasty Electrophysiologic study
RESULTS Of the 125,892 patients, 7789 (6.2%) received an implantable defibrillator within 90 days of hospital admission. Recipients were more likely to be younger, white, male, and urban dwelling; to have ischemic heart disease or heart failure; and to be admitted with a diagnosis of ventricular fibrillation to hospitals that were capable of coronary bypass surgery and defibrillator implantation procedures (Table 1). These patients were also more likely to receive other interventional procedures, includ-
71 ⫾ 5 97 17 75
76 ⫾ 7 93 40 72
20 60 32 17 17 13 13 16 5
12 39 25 21 15 13 16 10 4
6 19 8 16
7 8 4 9
12 13 6 12 8 3 2 3 46
13 16 5 12 13 2 4 3 49
23 59 43
28 42 17
63 13 3 77
17 2 1 12
ing angiography, coronary angioplasty, and coronary bypass surgery, within 90 days of initial admission. Mortality was lower in patients who received a defibrillator: 11% versus 23% at 1 year (OR ⫽ 0.43; 95% CI: 0.39 to 0.46), 21% versus 34% at 2 years (OR ⫽ 0.50; 95% CI: 0.47 to 0.54), and 29% versus 43% at 3 years (OR ⫽ 0.55; 95% CI: 0.51 to 0.59). Survival of the entire patient cohort improved significantly for patients enrolled more recently. The improvement was most evident at 1 year of follow-up and was sustained for a longer time (Figure 1). May 2002
THE AMERICAN JOURNAL OF MEDICINE威
Volume 112 521
Effectiveness and Cost-effectiveness of Implantable Defibrillators/Weiss et al
Figure 1. One-year, 2-year, and 3-year survival for the entire cohort by year of entry (n ⫽ 125 892).
The trend toward improved survival over time was statistically significant for the defibrillator recipients, but not for those who did not receive a defibrillator.
Propensity Score Model This analysis yielded a C-statistic (area under the receiver operating characteristic curve) of 0.84, indicating good discrimination between patients treated with and without a defibrillator. The strongest predictors of defibrillator placement were primary diagnosis of ventricular fibrillation, admission to a hospital with defibrillator implantation capability, and a history of prior myocardial infarction, ischemic heart disease, or cardiomyopathy. Patients who were aged ⬎90 years, aged between 85 and 90 years, black, or female were less likely to receive a defibrillator (Figure 2).
Matched-Pair Analysis The matching procedure resulted in 7612 pairs with equal overall probability of receiving a defibrillator based on all of the variables included in the propensity score model. The baseline characteristics between matched pairs were similar (Table 2), although patients who received a defibrillator were more likely to have undergone subsequent coronary angiography and coronary bypass surgery within 90 days of hospitalization. These patients also had significantly lower mortality than did the matched patients who did not receive a defibrillator: 11% versus 19% at 1 year (OR ⫽ 0.57; 95% CI: 0.51 to 0.63), 20% versus 30% at 2 years (OR ⫽ 0.66; 95% CI: 0.60 to 0.72), and 522
May 2002
THE AMERICAN JOURNAL OF MEDICINE威
Volume 112
28% versus 39% at 3 years (OR ⫽ 0.70; 95% CI: 0.63 to 0.77). The number of deaths was 710 versus 1133 at 1 year, 986 versus 1303 at 2 years, and 1074 versus 1358 at 3 years. The higher survival among these recipients persisted during the 8 years of follow-up (P ⫽ 0.0001), but the survival difference between the two patient groups decreased over time (Figure 3). Median undiscounted survival in the defibrillator group was 5.7 years (95% CI: 5.5 to 6.0 years) compared with 4.6 years (95% CI: 4.3 to 4.8 years) in the nondefibrillator group. The subgroup analysis of the 1269 matched pairs representing the middle tertile of the propensity score confirmed a significantly lower mortality among patients who received a defibrillator compared with those who did not; mortality was 14% versus 18% at 1 year (OR ⫽ 0.75; 95% CI: 0.60 to 0.93), 21% versus 28% at 2 years (OR ⫽ 0.69; 95% CI: 0.56 to 0.84), and 29% versus 37% at 3 years (OR ⫽ 0.70; 95% CI: 0.57 to 0.85). The number of deaths was 164 versus 209 at 1 year, 220 versus 288 at 2 years, and 265 versus 334 at 3 years.
Expenditures and Cost-effectiveness
Expenditures within the first year were $48,700 ⫾ $20,500 for the defibrillator group and $17,000 ⫾ $18,500 for the nondefibrillator group in the matchedpairs cohort (Figure 4). This difference of about $32,000 increased during the 8 years of follow-up to a final difference of $41,500 in discounted cumulative expenditures ($78,700 for the defibrillator group vs. $37,200 for the
Effectiveness and Cost-effectiveness of Implantable Defibrillators/Weiss et al
Figure 2. Independent predictors of receiving an implantable cardioverter defibrillator (ICD) in the propensity score model. Points correspond with the odds ratios for receiving a defibrillator. Bars indicate the 95% confidence intervals. Values ⬎1, indicating a higher likelihood of receiving an implantable defibrillator, lie on the right of the vertical line. Values ⬍1, indicating a lower likelihood of receiving a defibrillator, are on the left. All values are statistically significant (P ⬍0.0001).
nondefibrillator group). Discounted mean survival was 4.6 years for the defibrillator group and 4.1 years for the nondefibrillator group. The resulting marginal costeffectiveness of defibrillator vs. nondefibrillator treatment was $78,400 per life-year gained during 8 years of follow-up. At 3 years, the marginal cost-effectiveness was $133,500 per life-year gained.
DISCUSSION This study demonstrates that use of an implantable cardioverter defibrillator is associated with a significant improvement in survival in a large unselected cohort of older patients hospitalized for ventricular tachycardia or ventricular fibrillation. Our findings confirm the results of previous randomized trials and extend them by examining a less selected patient population in wider practice settings, and with a longer follow-up. Randomized controlled trials are accepted as the reference standard for comparison of medical therapies. However, the generalizability of results from such trials is limited by the use of stringent inclusion and exclusion criteria. Furthermore, randomized trials of devices and procedures are limited by the selection of sites with considerable experience and superior procedural outcomes. In our study, we adjusted for baseline differences between patients with ventricular tachycardia or ventricular
fibrillation who received an implantable defibrillator and those who were treated medically. The propensity score allowed us to assess the likelihood that each patient would receive a defibrillator based on the clinical and demographic descriptors available in the Medicare claims files. By matching each defibrillator recipient with a control patient who had an equal probability of receiving a defibrillator, we sought to create a cohort with equivalent severity of illness and prognosis. Although some residual differences remained in observed characteristics between groups, the small magnitude of these differences (a few percentage points) suggests that they are unlikely to be of clinical significance. We found that mortality among the defibrillator recipients was significantly lower than in clinically similar medically treated patients. The mortality reduction (34%) was similar to that seen in the Antiarrhythmics Versus Implantable Defibrillators trial (4), which reported a mortality reduction of 33% at 2 years, and in the Canadian Implantable Defibrillator Study (5), which reported a reduction of 35%. This concordance suggests that our observational analysis identified an actual therapeutic effect of the implantable defibrillator rather than a residual selection bias for defibrillator implantation. Conversely, the mortality reduction seen in our patients who had been treated across a broad spectrum of hospitals and regions suggests that the results of the randomMay 2002
THE AMERICAN JOURNAL OF MEDICINE威
Volume 112 523
Effectiveness and Cost-effectiveness of Implantable Defibrillators/Weiss et al
Table 2. Characteristics of the Matched Pairs Patients Who Received a Defibrillator (n ⫽ 7612)
Variable
Patients Who Received Medical Treatment (n ⫽ 7612)
P Value
Mean ⫾ SD or Percentage Demographic characteristics Age (years) White race Female sex Urban residence Cardiac comorbid conditions Primary diagnosis of ventricular fibrillation Ischemic heart disease Heart failure Hypertension Atrial fibrillation or flutter Valvular disease Cardiac conduction disorder Cardiomyopathy Acute myocardial infarction admission in prior year Heart failure admission in prior year Prior myocardial infarction Prior coronary angiography Prior bypass surgery Other comorbid conditions Fluid or electrolyte imbalance Pulmonary disease Central nervous system disorders Diabetes Syncope Respiratory failure or arrest Stroke Peripheral vascular disease Any hospitalization in prior year Hospital characteristics Angiography capability only Electrophysiologic study/implantable defibrillator capability Procedures within 90 days Angiography Bypass surgery Coronary angioplasty Electrophysiologic study
ized trials are generalizable beyond the narrow range of patients and providers that participated. One interesting observation was that the survival advantage of defibrillator recipients relative to medically treated patients narrowed substantially with longer follow-up, which was similar to that reported in the Cardiac Arrest Study Hamburg (6) that reported equal survival at 7 years (vs. 8 years in our study). It may be that the implantable defibrillator does not confer a long-lasting survival benefit, either because of disease progression in surviving patients or because of late device failures, revi524
May 2002
THE AMERICAN JOURNAL OF MEDICINE威
Volume 112
71 ⫾ 5 97 17 75
72 ⫾ 5 99 16 85
0.0001 0.0001 0.03 0.0001
20 60 32 17 17 13 13 16 6 6 19 8 16
18 60 31 17 17 13 13 15 6 6 20 7 16
0.0001 0.97 0.42 0.73 0.81 0.74 0.48 0.29 0.13 0.78 0.03 0.0002 0.83
12 13 6 12 8 3 2 3 46
10 12 5 13 9 3 2 3 46
0.0004 0.34 0.003 0.23 0.16 0.82 0.42 0.32 0.61
23 43
28 41
0.0001 0.003
63 13 3 77
28 5 3 29
0.0001 0.0001 0.96 0.0001
sions, and replacements. It is important to note, however, that by necessity the longest follow-up was derived from patients treated with earlier model devices. The effectiveness and reliability of implantable defibrillators are improving, so better clinical results might be expected with the use of more advanced devices. Indeed, the outcome in this cohort improved with time. Furthermore, the relatively small number of patients remaining at risk late in follow-up limits the statistical power of these observations. Nevertheless, data from Medicare and the Cardiac Arrest Study Hamburg trial underscore the importance
Effectiveness and Cost-effectiveness of Implantable Defibrillators/Weiss et al
Figure 3. Kaplan-Meier survival curves for the matched-pairs cohort. ICD ⫽ implantable cardioverter defibrillator.
of documenting long-term outcomes in both clinical trials and registries and suggest that the early benefits of the devices may not be sustained. Medicare patients treated with an implantable defibrillator had considerably higher expenditures for inpatient
medical care over long-term follow-up. The higher initial cost of treatment with a defibrillator is not surprising, because the device costs more than $20,000, and there are additional costs of implantation. A short-term perspective on costs can be misleading because therapies that are
Figure 4. Cumulative expenditures for the matched-pairs cohort, in 1999 U.S. dollars. ICD ⫽ implantable cardioverter defibrillator. May 2002
THE AMERICAN JOURNAL OF MEDICINE威
Volume 112 525
Effectiveness and Cost-effectiveness of Implantable Defibrillators/Weiss et al
Table 3. Estimated Cost Per Life-Year Saved for Various Medical Interventions Cost Per Life-Year Saved ($)
Intervention Beta-blockers for low-risk myocardial infarction survivors 3-Vessel bypass versus percutaneous intervention for severe angina Lovastatin for men aged 45 to 54 years with no heart disease and cholesterol level ⱖ300 mg/dL Dialysis for end-stage renal disease Heart transplantation for patients aged 50 years with terminal heart disease Annual mammography for women aged 55 to 64 years
17,000 23,000 34,000 51,000 100,000 110,000
From Tengs et al. (16).
initially more expensive may “pay for themselves” by reducing later treatment costs. However, the Medicare data showed that the cost differential increased with extended follow-up, as have other studies that reported persistently higher costs among defibrillator recipients (14,15). Whether the outcomes are sufficiently better to justify the added expense can be examined by cost-effectiveness analysis. When we compared the difference in discounted cumulative expenditures between defibrillator recipients and medically treated patients ($41,600) and the difference in discounted life-years saved (mean, 0.53 lifeyears), we obtained a cost-effectiveness ratio of about $78,000 per life-year added. This figure is higher than the benchmark of approximately $50,000 per life-year added (Table 3), which is derived from the willingness of society to pay for renal dialysis, although it falls within a “gray” zone of cost-effectiveness ratios between $50,000 and $100,000 per life-year saved, along with several other commonly accepted medical interventions (16). The cost-effectiveness ratio in our study was less favorable ($78,000 vs. $27,000 per life-year gained) than that in the Multicenter Automatic Defibrillator Implantation Trial (14), and closer to the $54,000 per life-year gained using the decision model by Owens et al. (17) under the assumptions of a lifetime follow-up and a 20% risk reduction due to the implantable defibrillator (17). Our findings were more favorable than the figures reported by the Canadian Implantable Defibrillator Study investigators ($138,800 per life-year gained) and the preliminary results of the Antiarrhythmics Versus Implantable Defibrillators Investigators (15). The latter studies, however, based their analyses on limited patient follow-up, which may underestimate the years of life added due to use of a defibrillator. Indeed, when our analysis was limited to a shorter follow-up, the cost-effectiveness ratio (about $134,000 per life-year gained at 3 years’ follow-up) was less favorable. Similarly, extrapolation of follow-up to 12 years by the Canadian Implantable Defibrillator Study investigators resulted in a more acceptable range of about $65,000 to $98,000 per life-year gained (15). The implantable defibrillator may therefore be economically ac526
May 2002
THE AMERICAN JOURNAL OF MEDICINE威
Volume 112
ceptable when used as secondary prevention for patients who have experienced an episode of sustained ventricular tachycardia or ventricular fibrillation. The major limitation of this study was its use of administrative claims data that lacked details about important clinical predictors of outcome, such as left ventricular ejection fraction and extent of coronary disease. Consequently, the statistical adjustment for baseline differences between patients could not be as complete as when using detailed clinical data. Although we matched patients closely using the propensity score method, our study was not randomized, and residual differences between patient groups may have affected the results. Finally, our cost data were derived from Medicare expenditures rather than a detailed cost accounting procedure. Any differences between expenditures and costs, however, would have affected all patients, and thus our analysis of relative costs between groups is less likely to have been affected. Outcomes among Medicare beneficiaries suggest that use of the implantable cardioverter defibrillator reduces mortality among unselected patients in typical practice settings. The cost of care for defibrillator treated patients is consistently higher than for medically treated patients, yet the higher costs may be acceptable in light of the improved survival in patients with ventricular tachycardia or ventricular fibrillation.
REFERENCES 1. American Heart Association. 2001 Heart and Stroke Statistical Update. Dallas, Texas: American Heart Association; 2000. 2. Murphy SL. Deaths: final data for 1998. National Vital Statistics Reports. Vol 48, No. 11. Hyattsville, Maryland: National Center for Health Statistics; 2000. 3. Myerburg RJ, Interian A Jr, Mitrani RM, et al. Frequency of sudden cardiac death and profiles of risk. Am J Cardiol. 1997;80:10F–19F. 4. The Antiarrhythmics Versus Implantable Defibrillators (AVID) Investigators. A comparison of antiarrhythmic-drug therapy with implantable defibrillators in patients resuscitated from near-fatal ventricular arrhythmias. N Engl J Med. 1997;337:1576 –1583. 5. Connolly SJ, Gent M, Roberts RS, et al. Canadian Implantable Defibrillator Study (CIDS): a randomized trial of the implantable car-
Effectiveness and Cost-effectiveness of Implantable Defibrillators/Weiss et al
6.
7.
8.
9.
10.
11.
dioverter defibrillator against amiodarone. Circulation. 2000;101: 1297–1302. Kuck KH, Cappato R, Siebels J, Ruppel R. Randomized comparison of antiarrhythmic drug therapy with implantable defibrillator in patients resuscitated from cardiac arrest: the Cardiac Arrest Study Hamburg (CASH). Circulation. 2000;102:748 –754. Udvarhelyi IS, Gatsonis C, Epstein AM, et al. Acute myocardial infarction in the Medicare population. Process of care and clinical outcomes. JAMA. 1992;268:2530 –2536. Pashos CL, Newhouse JP, McNeil BJ. Temporal changes in the care and outcomes of elderly patients with acute myocardial infarction, 1987 through 1990. JAMA. 1993;270:1832–1836. McClellan M, McNeil BJ, Newhouse JP. Does more intensive treatment of acute myocardial infarction in the elderly reduce mortality? Analysis using instrumental variables. JAMA. 1994;272:859 –866. Rosenbaum PR, Rubin DB. The central role of the propensity score in observational studies for causal effects. Biometrika. 1983;70:41– 55. Rosenbaum PR, Rubin DB. Reducing bias in observational studies using subclassification on the propensity score. J Am Stat Assoc. 1984;79:516 –524.
12. Connors AF Jr, Speroff T, Dawson NV, et al. The effectiveness of right heart catheterization in the initial care of critically ill patients. SUPPORT Investigators. JAMA. 1996;276:889 –897. 13. Etzioni R, Urban N, Baker M. Estimating the costs attributable to a disease with application to ovarian cancer. J Clin Epidemiol. 1996; 49:95–103. 14. Mushlin AI, Hall WJ, Zwanziger J, et al. The cost-effectiveness of automatic implantable cardiac defibrillators: results from MADIT. Multicenter Automatic Defibrillator Implantation Trial. Circulation. 1998;97:2129 –2135. 15. O’Brien BJ, Connolly SJ, Goeree R, et al. Cost-effectiveness of the implantable cardioverter-defibrillator: results from the Canadian Implantable Defibrillator Study (CIDS). Circulation. 2001;103: 1416 –1421. 16. Tengs TO, Adams ME, Pliskin JS, et al. Five hundred life-saving interventions and their cost-effectiveness. Risk Anal. 1995;15:369 – 389. 17. Owens DK, Sanders GD, Harris RA, et al. Cost-effectiveness of implantable cardioverter defibrillators relative to amiodarone for prevention of sudden cardiac death. Ann Intern Med. 1997;126:1–12.
May 2002
THE AMERICAN JOURNAL OF MEDICINE威
Volume 112 527
Effectiveness and Cost-effectiveness of Implantable Cardioverter Defibrillators in the Treatment of Ventricular Arrhythmias among Medicare Beneficiaries J. Peter Weiss, MD, Olga Saynina, MA, MBA, Kathryn M. McDonald, MM, Mark B. McClellan, MD, PhD, Mark A. Hlatky, MD PURPOSE: The implantable cardioverter defibrillator has been assessed in randomized trials, but the generalizability of trial results to broader clinical settings is unclear. Our purpose was to evaluate the outcomes and costs of defibrillator use in an unselected population. SUBJECTS AND METHODS: We identified 125,892 Medicare patients who were discharged between 1987 and 1995 after hospitalization with a primary diagnosis of ventricular tachycardia or ventricular fibrillation, 7789 of whom (6.2%) received a defibrillator. We used a multivariable propensity score that included patient and hospital characteristics to match pairs of patients, in which one patient received a defibrillator and the other did not. We compared mortality and costs in these 7612 matched pairs during 8 years of follow-up. RESULTS: Patients who received a defibrillator were more likely to be younger, white, male, and urban dwelling, and to have ischemic heart disease, heart failure, or a history of ven-
tricular fibrillation. In the matched-pairs analysis, those who received a defibrillator had significantly lower mortality: 11% versus 19% at 1 year (odds ratio [OR] ⫽ 0.57; 95% confidence interval [CI]: 0.51 to 0.63), 20% versus 30% at 2 years (OR ⫽ 0.66; 95% CI: 0.60 to 0.72), and 28% versus 39% at 3 years (OR ⫽ 0.70; 95% CI: 0.63 to 0.77). These patients also had lower mortality at 8 years (P ⫽ 0.0001), although this advantage over patients who received medical treatment only decreased over time. Expenditures among defibrillator recipients were consistently higher, with a cost-effectiveness ratio of $78,400 per lifeyear gained. CONCLUSION: The use of implantable defibrillators was associated with significantly lower mortality and higher costs, whereas the cost-effectiveness was higher than many, but not all, generally accepted therapies. Am J Med. 2002;112:519 –527. 䉷2002 by Excerpta Medica, Inc.
C
26,000 implantations performed in the United States in 1998 (1). The effectiveness of the implantable defibrillator in reducing mortality has been shown in randomized clinical trials (4 – 6). However, the patients in these trials were carefully selected and were treated in medical centers that had facilities for performing specific cardiac procedures, and the results therefore may not be generalized to less selected patients in more typical practice settings. In view of these limitations, the expense associated with implantation and maintenance of the device, and the number of patients who could benefit from its use, we sought to assess the clinical outcomes and cost-effectiveness of the implantable defibrillator in a population-based cohort of Medicare patients.
ardiac arrest affects an estimated 1000 persons each day in the United States and leads to 220,000 deaths each year (1). Sudden death accounts for nearly a third of all cardiac deaths and for up to half of deaths related to ischemic heart disease (2). Patients who have had out-of-hospital cardiac arrest are particularly at high risk (3), but their outcomes have improved considerably in recent years, particularly with the development of the implantable cardioverter defibrillator. The implantable defibrillator is widely used, with approximately
From the Department of Health Research and Policy (JPW, MAH), Stanford, California; and Department of Medicine (KMM, MBM), Stanford University School of Medicine, Stanford, California; and the National Bureau of Economics Research (OS, MBM), Cambridge, Massachusetts. Supported by Grant HS 08362 from the Agency for Healthcare Research and Quality, Rockville, Maryland. Requests for reprints should be addressed to Mark A. Hlatky, MD, Stanford University School of Medicine, HRP Redwood Building, Room 150, Stanford, California 94305-5405. Manuscript submitted June 11, 2001, and accepted in revised form January 18, 2002. 䉷2002 by Excerpta Medica, Inc. All rights reserved.
MATERIALS AND METHODS We used the Health Care Financing Administration Medicare Provider Analysis and Review inpatient hospitalization file to identify all patients who were discharged 0002-9343/02/$–see front matter 519 PII S0002-9343(02)01078-1
Effectiveness and Cost-effectiveness of Implantable Defibrillators/Weiss et al
from a U.S. hospital between January 1, 1987, and September 30, 1995, and who had a primary discharge diagnosis of ventricular fibrillation or ventricular tachycardia (International Classification of Diseases, Ninth Revision, Clinical Modification [ICD-9-CM] codes 427.1, 427.4, 427.5). Patients who had been hospitalized for ventricular fibrillation or ventricular tachycardia between January 1, 1984, and December 31, 1986, were excluded. We began inclusion in 1987 because it was the first year in which at least 1% of the patients in the cohort received an implantable defibrillator. Recipients were defined as patients who received a device within 90 days of hospital admission (ICD-9-CM procedure codes 37.94 to 37.96); patients who underwent generator replacements or lead revisions were excluded. To ensure the stability of the cohort (7–9), we used data from the Medicare Beneficiary Health Insurance Skeletonized Eligibility Write-off file to exclude patients who were younger than 65 years; did not reside in the United States; were enrolled in a health maintenance organization; had noncontinuous enrollment in Medicare Part A or Part B during the 1 year following initial admission; were admitted to a non–acute care hospital, federal hospital, or hospital outside the United States; had invalid mortality data; or had missing Write-off file data. We excluded patients with possible miscoding of ventricular arrhythmia as the primary diagnosis: discharge with a total stay ⬍2 days, admittance on the same day as or the day after discharge from an admission unrelated to ventricular arrhythmia, acute myocardial infarction during index admission, or elective admission. Elective defibrillator implantations that were thought to be unrelated to the index ventricular arrhythmia event were eliminated by excluding patients who received an implant on the day of index admission or the day after admission. The cohort was limited to those discharged alive from the index admission to avoid bias due to the selection of therapy in moribund patients. These exclusions were not mutually exclusive and resulted in a final cohort of 125,892 patients. A single longitudinal record for each patient was created by linking the index admission to all previous and subsequent admissions. Expenditures were defined as the sum of diagnostic related group payments plus outlier payments, normalized to 1999 U.S. dollars using the Gross Domestic Product Deflator inflation index (http:// www.jsc.nasa.gov/bu2/inflateGDP.html).
Statistical Analysis Unadjusted analysis. All-cause mortality at 1, 2, and 3 years was compared between patients who did and did not receive a defibrillator, with results reported as odds ratios (ORs) and 95% confidence intervals (CIs). Propensity score. To address the possibility that the primary outcomes might be biased by the selection of pa520
May 2002
THE AMERICAN JOURNAL OF MEDICINE威
Volume 112
tients for treatment with an implantable defibrillator, we performed a propensity score adjustment (10,11). Demographic and clinical data thought to be predictive of treatment selection were entered as independent variables in a multivariable logistic regression that had receipt of an implantable defibrillator as the dependent variable. All variables were retained in the model, regardless of the level of statistical significance. Demographic variables included sex, race, age (in 5-year intervals), and rural versus urban residence. Comorbid conditions and secondary diagnoses included admission with a primary diagnosis of ventricular fibrillation versus ventricular tachycardia, ischemic heart disease, heart failure, prior myocardial infarction, atrial fibrillation or flutter, cardiac conduction abnormality, valvular disease, cardiomyopathy, chronic obstructive pulmonary disease, fluid or electrolyte imbalance, hypertension, diabetes, syncope, peripheral vascular disease, respiratory failure or arrest, stroke, and other central nervous system disorders. Also included were year of entry, hospital admission within the previous year for heart failure or myocardial infarction, hospital admission within the previous year for any reason, and prior coronary bypass surgery or angioplasty. We adjusted for hospital characteristics such as capability for cardiac catheterization, cardiac surgery, electrophysiologic study, and defibrillator implantation. A propensity score was calculated for each patient that yielded the probability (0 to 1) of receipt of an implantable defibrillator, based on demographic, diagnostic, and hospital characteristics. Matched-pairs analysis. Patients who received a defibrillator were matched one to one with patients who did not receive a defibrillator but who had an equivalent propensity score (12), defined as an absolute value (propensity score for defibrillator patient minus propensity score for nondefibrillator patient) ⱕ0.01. Those who did not receive a defibrillator were accepted as matches only if their survival time was equal to or greater than the time to device implantation for the corresponding defibrillator patient. Mortality for the matched cohort at 1, 2, and 3 years was estimated using logistic regression, as well as 8-year Kaplan-Meier cumulative survival. Cumulative expenditures were calculated in 1999 constant dollars, discounted at 3% per year of follow-up (13). The cumulative expenditures and mean cumulative survival (discounted at 3% per year) were used to calculate marginal cost-effectiveness as (Costdefibrillator minus Costnondefibrillator) / (LifeYearsdefibrillator minus Life-Yearsnondefibrillator). One-, 2-, and 3-year mortality for a subset of the matched pairs within the middle tertile of propensity score was also examined using logistic regression to assess the outcomes of patients with an intermediate likelihood of receiving either course of therapy. Analysis was performed using SAS software, version 6.12 (SAS Institute, Cary, North Carolina).
Effectiveness and Cost-effectiveness of Implantable Defibrillators/Weiss et al
Table 1. Characteristics of the Entire Medicare Cohort
Variable
Patients Who Received Patients Who Received a Defibrillator Medical Treatment (n ⫽ 7789) (n ⫽ 118 103) Mean ⫾ SD or Percentage
Demographic characteristics Age (years) White race Female sex Urban residence Cardiac comorbid conditions Primary diagnosis of ventricular fibrillation Ischemic heart disease Heart failure Hypertension Atrial fibrillation or flutter Valvular disease Cardiac conduction disorder Cardiomyopathy Acute myocardial infarction admission in prior year Heart failure admission in prior year Prior myocardial infarction Prior coronary angioplasty Prior bypass surgery Other comorbid conditions Fluid or electrolyte imbalance Pulmonary disease Central nervous system disorders Diabetes Syncope Respiratory failure or arrest Stroke Peripheral vascular disease Any hospitalization in prior year Hospital characteristics Angiography capability only Bypass surgery capability Electrophysiologic study/implantable defibrillator capability Procedures performed within 90 days Angiography Bypass surgery Coronary angioplasty Electrophysiologic study
RESULTS Of the 125,892 patients, 7789 (6.2%) received an implantable defibrillator within 90 days of hospital admission. Recipients were more likely to be younger, white, male, and urban dwelling; to have ischemic heart disease or heart failure; and to be admitted with a diagnosis of ventricular fibrillation to hospitals that were capable of coronary bypass surgery and defibrillator implantation procedures (Table 1). These patients were also more likely to receive other interventional procedures, includ-
71 ⫾ 5 97 17 75
76 ⫾ 7 93 40 72
20 60 32 17 17 13 13 16 5
12 39 25 21 15 13 16 10 4
6 19 8 16
7 8 4 9
12 13 6 12 8 3 2 3 46
13 16 5 12 13 2 4 3 49
23 59 43
28 42 17
63 13 3 77
17 2 1 12
ing angiography, coronary angioplasty, and coronary bypass surgery, within 90 days of initial admission. Mortality was lower in patients who received a defibrillator: 11% versus 23% at 1 year (OR ⫽ 0.43; 95% CI: 0.39 to 0.46), 21% versus 34% at 2 years (OR ⫽ 0.50; 95% CI: 0.47 to 0.54), and 29% versus 43% at 3 years (OR ⫽ 0.55; 95% CI: 0.51 to 0.59). Survival of the entire patient cohort improved significantly for patients enrolled more recently. The improvement was most evident at 1 year of follow-up and was sustained for a longer time (Figure 1). May 2002
THE AMERICAN JOURNAL OF MEDICINE威
Volume 112 521
Effectiveness and Cost-effectiveness of Implantable Defibrillators/Weiss et al
Figure 1. One-year, 2-year, and 3-year survival for the entire cohort by year of entry (n ⫽ 125 892).
The trend toward improved survival over time was statistically significant for the defibrillator recipients, but not for those who did not receive a defibrillator.
Propensity Score Model This analysis yielded a C-statistic (area under the receiver operating characteristic curve) of 0.84, indicating good discrimination between patients treated with and without a defibrillator. The strongest predictors of defibrillator placement were primary diagnosis of ventricular fibrillation, admission to a hospital with defibrillator implantation capability, and a history of prior myocardial infarction, ischemic heart disease, or cardiomyopathy. Patients who were aged ⬎90 years, aged between 85 and 90 years, black, or female were less likely to receive a defibrillator (Figure 2).
Matched-Pair Analysis The matching procedure resulted in 7612 pairs with equal overall probability of receiving a defibrillator based on all of the variables included in the propensity score model. The baseline characteristics between matched pairs were similar (Table 2), although patients who received a defibrillator were more likely to have undergone subsequent coronary angiography and coronary bypass surgery within 90 days of hospitalization. These patients also had significantly lower mortality than did the matched patients who did not receive a defibrillator: 11% versus 19% at 1 year (OR ⫽ 0.57; 95% CI: 0.51 to 0.63), 20% versus 30% at 2 years (OR ⫽ 0.66; 95% CI: 0.60 to 0.72), and 522
May 2002
THE AMERICAN JOURNAL OF MEDICINE威
Volume 112
28% versus 39% at 3 years (OR ⫽ 0.70; 95% CI: 0.63 to 0.77). The number of deaths was 710 versus 1133 at 1 year, 986 versus 1303 at 2 years, and 1074 versus 1358 at 3 years. The higher survival among these recipients persisted during the 8 years of follow-up (P ⫽ 0.0001), but the survival difference between the two patient groups decreased over time (Figure 3). Median undiscounted survival in the defibrillator group was 5.7 years (95% CI: 5.5 to 6.0 years) compared with 4.6 years (95% CI: 4.3 to 4.8 years) in the nondefibrillator group. The subgroup analysis of the 1269 matched pairs representing the middle tertile of the propensity score confirmed a significantly lower mortality among patients who received a defibrillator compared with those who did not; mortality was 14% versus 18% at 1 year (OR ⫽ 0.75; 95% CI: 0.60 to 0.93), 21% versus 28% at 2 years (OR ⫽ 0.69; 95% CI: 0.56 to 0.84), and 29% versus 37% at 3 years (OR ⫽ 0.70; 95% CI: 0.57 to 0.85). The number of deaths was 164 versus 209 at 1 year, 220 versus 288 at 2 years, and 265 versus 334 at 3 years.
Expenditures and Cost-effectiveness
Expenditures within the first year were $48,700 ⫾ $20,500 for the defibrillator group and $17,000 ⫾ $18,500 for the nondefibrillator group in the matchedpairs cohort (Figure 4). This difference of about $32,000 increased during the 8 years of follow-up to a final difference of $41,500 in discounted cumulative expenditures ($78,700 for the defibrillator group vs. $37,200 for the
Effectiveness and Cost-effectiveness of Implantable Defibrillators/Weiss et al
Figure 2. Independent predictors of receiving an implantable cardioverter defibrillator (ICD) in the propensity score model. Points correspond with the odds ratios for receiving a defibrillator. Bars indicate the 95% confidence intervals. Values ⬎1, indicating a higher likelihood of receiving an implantable defibrillator, lie on the right of the vertical line. Values ⬍1, indicating a lower likelihood of receiving a defibrillator, are on the left. All values are statistically significant (P ⬍0.0001).
nondefibrillator group). Discounted mean survival was 4.6 years for the defibrillator group and 4.1 years for the nondefibrillator group. The resulting marginal costeffectiveness of defibrillator vs. nondefibrillator treatment was $78,400 per life-year gained during 8 years of follow-up. At 3 years, the marginal cost-effectiveness was $133,500 per life-year gained.
DISCUSSION This study demonstrates that use of an implantable cardioverter defibrillator is associated with a significant improvement in survival in a large unselected cohort of older patients hospitalized for ventricular tachycardia or ventricular fibrillation. Our findings confirm the results of previous randomized trials and extend them by examining a less selected patient population in wider practice settings, and with a longer follow-up. Randomized controlled trials are accepted as the reference standard for comparison of medical therapies. However, the generalizability of results from such trials is limited by the use of stringent inclusion and exclusion criteria. Furthermore, randomized trials of devices and procedures are limited by the selection of sites with considerable experience and superior procedural outcomes. In our study, we adjusted for baseline differences between patients with ventricular tachycardia or ventricular
fibrillation who received an implantable defibrillator and those who were treated medically. The propensity score allowed us to assess the likelihood that each patient would receive a defibrillator based on the clinical and demographic descriptors available in the Medicare claims files. By matching each defibrillator recipient with a control patient who had an equal probability of receiving a defibrillator, we sought to create a cohort with equivalent severity of illness and prognosis. Although some residual differences remained in observed characteristics between groups, the small magnitude of these differences (a few percentage points) suggests that they are unlikely to be of clinical significance. We found that mortality among the defibrillator recipients was significantly lower than in clinically similar medically treated patients. The mortality reduction (34%) was similar to that seen in the Antiarrhythmics Versus Implantable Defibrillators trial (4), which reported a mortality reduction of 33% at 2 years, and in the Canadian Implantable Defibrillator Study (5), which reported a reduction of 35%. This concordance suggests that our observational analysis identified an actual therapeutic effect of the implantable defibrillator rather than a residual selection bias for defibrillator implantation. Conversely, the mortality reduction seen in our patients who had been treated across a broad spectrum of hospitals and regions suggests that the results of the randomMay 2002
THE AMERICAN JOURNAL OF MEDICINE威
Volume 112 523
Effectiveness and Cost-effectiveness of Implantable Defibrillators/Weiss et al
Table 2. Characteristics of the Matched Pairs Patients Who Received a Defibrillator (n ⫽ 7612)
Variable
Patients Who Received Medical Treatment (n ⫽ 7612)
P Value
Mean ⫾ SD or Percentage Demographic characteristics Age (years) White race Female sex Urban residence Cardiac comorbid conditions Primary diagnosis of ventricular fibrillation Ischemic heart disease Heart failure Hypertension Atrial fibrillation or flutter Valvular disease Cardiac conduction disorder Cardiomyopathy Acute myocardial infarction admission in prior year Heart failure admission in prior year Prior myocardial infarction Prior coronary angiography Prior bypass surgery Other comorbid conditions Fluid or electrolyte imbalance Pulmonary disease Central nervous system disorders Diabetes Syncope Respiratory failure or arrest Stroke Peripheral vascular disease Any hospitalization in prior year Hospital characteristics Angiography capability only Electrophysiologic study/implantable defibrillator capability Procedures within 90 days Angiography Bypass surgery Coronary angioplasty Electrophysiologic study
ized trials are generalizable beyond the narrow range of patients and providers that participated. One interesting observation was that the survival advantage of defibrillator recipients relative to medically treated patients narrowed substantially with longer follow-up, which was similar to that reported in the Cardiac Arrest Study Hamburg (6) that reported equal survival at 7 years (vs. 8 years in our study). It may be that the implantable defibrillator does not confer a long-lasting survival benefit, either because of disease progression in surviving patients or because of late device failures, revi524
May 2002
THE AMERICAN JOURNAL OF MEDICINE威
Volume 112
71 ⫾ 5 97 17 75
72 ⫾ 5 99 16 85
0.0001 0.0001 0.03 0.0001
20 60 32 17 17 13 13 16 6 6 19 8 16
18 60 31 17 17 13 13 15 6 6 20 7 16
0.0001 0.97 0.42 0.73 0.81 0.74 0.48 0.29 0.13 0.78 0.03 0.0002 0.83
12 13 6 12 8 3 2 3 46
10 12 5 13 9 3 2 3 46
0.0004 0.34 0.003 0.23 0.16 0.82 0.42 0.32 0.61
23 43
28 41
0.0001 0.003
63 13 3 77
28 5 3 29
0.0001 0.0001 0.96 0.0001
sions, and replacements. It is important to note, however, that by necessity the longest follow-up was derived from patients treated with earlier model devices. The effectiveness and reliability of implantable defibrillators are improving, so better clinical results might be expected with the use of more advanced devices. Indeed, the outcome in this cohort improved with time. Furthermore, the relatively small number of patients remaining at risk late in follow-up limits the statistical power of these observations. Nevertheless, data from Medicare and the Cardiac Arrest Study Hamburg trial underscore the importance
Effectiveness and Cost-effectiveness of Implantable Defibrillators/Weiss et al
Figure 3. Kaplan-Meier survival curves for the matched-pairs cohort. ICD ⫽ implantable cardioverter defibrillator.
of documenting long-term outcomes in both clinical trials and registries and suggest that the early benefits of the devices may not be sustained. Medicare patients treated with an implantable defibrillator had considerably higher expenditures for inpatient
medical care over long-term follow-up. The higher initial cost of treatment with a defibrillator is not surprising, because the device costs more than $20,000, and there are additional costs of implantation. A short-term perspective on costs can be misleading because therapies that are
Figure 4. Cumulative expenditures for the matched-pairs cohort, in 1999 U.S. dollars. ICD ⫽ implantable cardioverter defibrillator. May 2002
THE AMERICAN JOURNAL OF MEDICINE威
Volume 112 525
Effectiveness and Cost-effectiveness of Implantable Defibrillators/Weiss et al
Table 3. Estimated Cost Per Life-Year Saved for Various Medical Interventions Cost Per Life-Year Saved ($)
Intervention Beta-blockers for low-risk myocardial infarction survivors 3-Vessel bypass versus percutaneous intervention for severe angina Lovastatin for men aged 45 to 54 years with no heart disease and cholesterol level ⱖ300 mg/dL Dialysis for end-stage renal disease Heart transplantation for patients aged 50 years with terminal heart disease Annual mammography for women aged 55 to 64 years
17,000 23,000 34,000 51,000 100,000 110,000
From Tengs et al. (16).
initially more expensive may “pay for themselves” by reducing later treatment costs. However, the Medicare data showed that the cost differential increased with extended follow-up, as have other studies that reported persistently higher costs among defibrillator recipients (14,15). Whether the outcomes are sufficiently better to justify the added expense can be examined by cost-effectiveness analysis. When we compared the difference in discounted cumulative expenditures between defibrillator recipients and medically treated patients ($41,600) and the difference in discounted life-years saved (mean, 0.53 lifeyears), we obtained a cost-effectiveness ratio of about $78,000 per life-year added. This figure is higher than the benchmark of approximately $50,000 per life-year added (Table 3), which is derived from the willingness of society to pay for renal dialysis, although it falls within a “gray” zone of cost-effectiveness ratios between $50,000 and $100,000 per life-year saved, along with several other commonly accepted medical interventions (16). The cost-effectiveness ratio in our study was less favorable ($78,000 vs. $27,000 per life-year gained) than that in the Multicenter Automatic Defibrillator Implantation Trial (14), and closer to the $54,000 per life-year gained using the decision model by Owens et al. (17) under the assumptions of a lifetime follow-up and a 20% risk reduction due to the implantable defibrillator (17). Our findings were more favorable than the figures reported by the Canadian Implantable Defibrillator Study investigators ($138,800 per life-year gained) and the preliminary results of the Antiarrhythmics Versus Implantable Defibrillators Investigators (15). The latter studies, however, based their analyses on limited patient follow-up, which may underestimate the years of life added due to use of a defibrillator. Indeed, when our analysis was limited to a shorter follow-up, the cost-effectiveness ratio (about $134,000 per life-year gained at 3 years’ follow-up) was less favorable. Similarly, extrapolation of follow-up to 12 years by the Canadian Implantable Defibrillator Study investigators resulted in a more acceptable range of about $65,000 to $98,000 per life-year gained (15). The implantable defibrillator may therefore be economically ac526
May 2002
THE AMERICAN JOURNAL OF MEDICINE威
Volume 112
ceptable when used as secondary prevention for patients who have experienced an episode of sustained ventricular tachycardia or ventricular fibrillation. The major limitation of this study was its use of administrative claims data that lacked details about important clinical predictors of outcome, such as left ventricular ejection fraction and extent of coronary disease. Consequently, the statistical adjustment for baseline differences between patients could not be as complete as when using detailed clinical data. Although we matched patients closely using the propensity score method, our study was not randomized, and residual differences between patient groups may have affected the results. Finally, our cost data were derived from Medicare expenditures rather than a detailed cost accounting procedure. Any differences between expenditures and costs, however, would have affected all patients, and thus our analysis of relative costs between groups is less likely to have been affected. Outcomes among Medicare beneficiaries suggest that use of the implantable cardioverter defibrillator reduces mortality among unselected patients in typical practice settings. The cost of care for defibrillator treated patients is consistently higher than for medically treated patients, yet the higher costs may be acceptable in light of the improved survival in patients with ventricular tachycardia or ventricular fibrillation.
REFERENCES 1. American Heart Association. 2001 Heart and Stroke Statistical Update. Dallas, Texas: American Heart Association; 2000. 2. Murphy SL. Deaths: final data for 1998. National Vital Statistics Reports. Vol 48, No. 11. Hyattsville, Maryland: National Center for Health Statistics; 2000. 3. Myerburg RJ, Interian A Jr, Mitrani RM, et al. Frequency of sudden cardiac death and profiles of risk. Am J Cardiol. 1997;80:10F–19F. 4. The Antiarrhythmics Versus Implantable Defibrillators (AVID) Investigators. A comparison of antiarrhythmic-drug therapy with implantable defibrillators in patients resuscitated from near-fatal ventricular arrhythmias. N Engl J Med. 1997;337:1576 –1583. 5. Connolly SJ, Gent M, Roberts RS, et al. Canadian Implantable Defibrillator Study (CIDS): a randomized trial of the implantable car-
Effectiveness and Cost-effectiveness of Implantable Defibrillators/Weiss et al
6.
7.
8.
9.
10.
11.
dioverter defibrillator against amiodarone. Circulation. 2000;101: 1297–1302. Kuck KH, Cappato R, Siebels J, Ruppel R. Randomized comparison of antiarrhythmic drug therapy with implantable defibrillator in patients resuscitated from cardiac arrest: the Cardiac Arrest Study Hamburg (CASH). Circulation. 2000;102:748 –754. Udvarhelyi IS, Gatsonis C, Epstein AM, et al. Acute myocardial infarction in the Medicare population. Process of care and clinical outcomes. JAMA. 1992;268:2530 –2536. Pashos CL, Newhouse JP, McNeil BJ. Temporal changes in the care and outcomes of elderly patients with acute myocardial infarction, 1987 through 1990. JAMA. 1993;270:1832–1836. McClellan M, McNeil BJ, Newhouse JP. Does more intensive treatment of acute myocardial infarction in the elderly reduce mortality? Analysis using instrumental variables. JAMA. 1994;272:859 –866. Rosenbaum PR, Rubin DB. The central role of the propensity score in observational studies for causal effects. Biometrika. 1983;70:41– 55. Rosenbaum PR, Rubin DB. Reducing bias in observational studies using subclassification on the propensity score. J Am Stat Assoc. 1984;79:516 –524.
12. Connors AF Jr, Speroff T, Dawson NV, et al. The effectiveness of right heart catheterization in the initial care of critically ill patients. SUPPORT Investigators. JAMA. 1996;276:889 –897. 13. Etzioni R, Urban N, Baker M. Estimating the costs attributable to a disease with application to ovarian cancer. J Clin Epidemiol. 1996; 49:95–103. 14. Mushlin AI, Hall WJ, Zwanziger J, et al. The cost-effectiveness of automatic implantable cardiac defibrillators: results from MADIT. Multicenter Automatic Defibrillator Implantation Trial. Circulation. 1998;97:2129 –2135. 15. O’Brien BJ, Connolly SJ, Goeree R, et al. Cost-effectiveness of the implantable cardioverter-defibrillator: results from the Canadian Implantable Defibrillator Study (CIDS). Circulation. 2001;103: 1416 –1421. 16. Tengs TO, Adams ME, Pliskin JS, et al. Five hundred life-saving interventions and their cost-effectiveness. Risk Anal. 1995;15:369 – 389. 17. Owens DK, Sanders GD, Harris RA, et al. Cost-effectiveness of implantable cardioverter defibrillators relative to amiodarone for prevention of sudden cardiac death. Ann Intern Med. 1997;126:1–12.
May 2002
THE AMERICAN JOURNAL OF MEDICINE威
Volume 112 527