Disparities in Use of Implantable Cardioverter-Defibrillators: Title and subTitle BreakMoving Beyond Process Measures to Outcomes Data

In recent years, significant disparities have been increasingly recognized among different patient populations in the health care system—disparities that largely mirror those in other aspects of society. An Institute of Medicine report from 2002 highlighted these disparities, finding that a consistent body of research shows substantial variations in the rates of medical procedures by race/ethnicity, even when insurance status, income, age, and severity of conditions are comparable.¹ Moreover, it has been known for many years that women and racial/ethnic minorities are less likely to be offered procedures such as cardiac catheterization, coronary artery bypass graft surgery, and carotid endarterectomy.^{2 - 4}

Implantable cardioverter-defibrillators (ICDs), first introduced about 20 years ago, are widely recognized as a potentially lifesaving therapy for those at highest risk for sudden cardiac death. The use of ICDs has increased steadily since the results of 2 clinical trials—the Multicenter Automatic Defibrillator Implantation Study (MADIT-II)⁵ and Sudden Cardiac Death in Heart Failure trial (SCD-HeFT)⁶ —showed that ICDs provide a mortality benefit when used for primary prevention. In late 2004, Medicare criteria for coverage of ICD for primary prevention was greatly expanded to include patients with New York Heart Association class II or III heart failure and left ventricular ejection fraction less than 35%.⁷

In this issue of JAMA, 2 studies provide important information about who is and is not receiving ICDs. Curtis and colleagues⁸ analyzed data from a 5% national sample of Medicare beneficiaries from 1999-2005 and found that among male Medicare beneficiaries in 2005 who met criteria for primary prevention, 32.3/1000 received a defibrillator within 1 year of diagnosis. However, defibrillators were implanted in only 8.6/1000 of such female beneficiaries. Thus, although the Centers for Medicare & Medicaid Services met its stated goal of 25 000 ICD implants in the first year after expanding coverage, these were performed mostly in men. Similarly, Hernandez and colleagues⁹ used data from the “Get with the Guidelines” Heart Failure database from 2005-2007 and found that 35.4% of patients who were eligible for ICD therapy received an implant by discharge. Again, women and racial/ethnic minorities were less likely than white men to have an ICD implanted. Compared with white men, black men were only 73% as likely to receive an ICD, whereas white women and black women were only 62% and 56% as likely to receive an ICD, respectively.

Thus, ICD implantation appears to be another in the list of procedures that are less likely to be provided to women and racial/ethnic minorities, and these data again highlight disturbing patterns of health care inequality. As with similar studies, insurance status does not appear related to this differential treatment and would not in any event explain the differences in the Medicare population.

Reasons that ICD implantation differs so much by sex and race/ethnicity are unclear. Perhaps, as the authors note, some of the differences are due to clinical factors. For example, women are more likely to have heart failure with preserved systolic function, which would not meet criteria for ICD implantation. Also women tend to be older and have more comorbidities when they are diagnosed with heart failure. To some extent, the differences in rates of ICD use simply may reflect the results of clinical trials, which included mostly white men and data to show benefit of ICD use in women. For example, MADIT-II (16% women) reported no benefit of ICD use in women (the study did not report statistics regarding black patients). Similarly, SCD-HeFT (23% women) found no benefit for ICD use among women (hazard ratio, 0.96). SCD-HeFT also found a nonsignificant benefit (hazard ratio, 0.75) for nonwhite patients (24% of study participants). It is also possible that women, minorities, or both refuse defibrillators at higher rates than white men do; indeed, patient refusal data for ICDs (and other invasive devices) would be helpful.

Having identified these disparities in ICD use, these reports lead to a much more fundamental question: Do lower ICD implantation rates in women and minorities lead to worse outcomes? Curtis et al⁸ provide the first data from “real world” experience to address this question and the results are troubling, but not for the expected reasons. After controlling for comorbidities, the authors found that Medicare beneficiaries who received ICD implants for primary prevention had no benefit in terms of reduced all-cause mortality. In other words, the bad news may not be for women and minorities, but for white men who are undergoing a procedure that, for primary prevention, has not been shown to extend their lives.

Why are these findings reported by Curtis et al different from the randomized clinical trial results? Patients in the clinical trials were younger and healthier than the Medicare population,¹⁰ and it may not be reasonable to expect the same benefit of ICD implants in an older, sicker population that has more competing causes of death. A recent analysis of 2467 ICD implants in Canada found that ICD recipients aged 65 through 74 years were twice as likely to die, and those 75 years or older were 3 times as likely to die as those younger than 65 years.¹¹ Similarly, patients with comorbidities had a lower mortality benefit and higher mortality risk associated with ICD implantation.

Curtis et al note that ICD therapy was not randomly assigned, and they suggest that Medicare beneficiaries who received ICDs may have been at higher risk than those who did not. However, the investigators adjusted for age, comorbidity, year of cohort entry, and probability of treatment, reducing the likelihood of this explanation. Moreover, it is at least as reasonable to assume that patients who were at higher risk of death due to comorbidities and older age would be less likely to receive ICDs. In either case, the Medicare data regarding mortality in patients who receive an ICD are discomforting.

Even if the benefits of ICDs on mortality are not confirmed, what about the influence of ICDs on quality of life? Unfortunately, the data here also are not encouraging. In a study based on interviews of MADIT-II participants, health-related quality of life (measured using the HU 13 score, a 0-1 utility scale describing health-related quality of life and considered an appropriate source of preferences for calculating quality-adjusted life-years for use in cost-effectiveness analyses¹² ) started low and improved (from 0.646 at baseline to 0.678 after 3 years) in those randomized to the conventional (non-ICD) group, but declined in the ICD group (from 0.637 to 0.601 after 3 years).¹³ The investigators state that “since we find no statistically significant quality-adjusted life-year [QALY] benefit. . . we concluded that the 3-year incremental cost-effectiveness ratio would fall above $235,000/QALY with 95% confidence.” In an earlier analysis, the Markov model for MADIT-II had reported a predicted incremental cost-effectiveness of $53 100, which assumed a 1.47 increase in QALY related to ICD and a cost of $27 975 for a single-chamber ICD.¹⁴ In actual practice, many patients have no quality of life benefit and receive more complex and expensive devices; for example, approximately 25% of patients in the SEARCH-MI primary prevention registry received a device for cardiac resynchronization therapy.¹⁵ Ultimately, the MADIT-II quality of life investigators concluded that adverse effects of the ICD on health-related quality of life, together with lower quality of life among survivors, may offset the 3-year survival benefits of ICDs.¹³

Thus, the multibillion-dollar question is: Are too few ICDs for primary prevention being implanted in women (and minorities) or are too many ICDs being implanted in (white) men? The important clinical and policy question may be not why women and black Medicare beneficiaries are less likely to get an ICD, but which Medicare beneficiaries will benefit from ICD at all? To answer this question, studies must look beyond reporting only process measures, such as implantation rates, and must include clinical outcomes, such as survival and quality of life after ICD implantation for primary and secondary prevention. By reporting the first outcomes data for ICD in the Medicare population, the study by Curtis et al should stimulate national dialogue on this crucial question.

Thus, in addition to their findings regarding disparities in ICD use, the studies by Curtis et al and Hernandez et al raise, perhaps inadvertently, a more serious concern. Their reports are important, but only first steps in understanding how to optimize delivery of cardiovascular health care in the United States. Their work highlights the importance of outcomes data for new therapies such as ICDs, reported according to sex and race/ethnicity subgroups, to determine if all patients are benefiting from health care advances.