In this regard, long-term dialysis, which is an intensive and expensive form of life support, is a good example to consider. The Medicare End-stage Renal Disease (ESRD) Program stands as the only federal program that finances disease-specific services to a segment of the US population on a near-universal basis. Under Medicare, persons diagnosed with ESRD who are eligible for Social Security receive coverage for most dialysis services and supplies. Patients aged 75 years or older currently represent one of the fastest growing groups within the ESRD population.³ Average Medicare costs for an older patient receiving long-term dialysis exceed $100 000 during the first year after initiation of therapy.⁴ Life expectancy on dialysis is often extremely limited for very elderly patients (ie, ≥85 years) and those with a higher burden of comorbidity and/or functional impairment.⁵ A substantial number of patients eventually discontinue dialysis, often in the setting of acute illness or failure to thrive. To date, little is known about treatment practices for older adults with ESRD and the extent to which these practices vary geographically. We describe the incidence of ESRD, the degree of preparedness for ESRD, and end-of-life care practices among older adults with ESRD across regions with differing intensities of care.

The primary predictor variable for all analyses was each hospital referral region’s end-of-life expenditure index, which is a measure of intensity of care during the last 6 months of life.¹ The index reflects both physician spending (from the Medicare Carrier File) and acute hospital spending (from the Medicare Provider Analysis and Review File) among Medicare beneficiaries who were between the ages of 65 and 100 years at the time of death. Only those who were eligible for Medicare for the 6-month period before death and were not enrolled in a health maintenance organization during this time were used to calculate the index. The index is calculated based on standardized national prices and is adjusted for the age, race, and sex of the Medicare beneficiaries in each hospital referral region. As such, it is intended to reflect that component of regional Medicare spending attributable to the overall quantity of medical services provided rather than to local differences in pricing and demographic structure. We obtained the most recent version of the end-of-life expenditure index (based on deaths occurring from 2000–2003) from the Dartmouth Atlas of Healthcare.⁶ Hospital referral regions were categorized by quintile of the end-of-life expenditure index.

Incident cases of treated ESRD (defined as first receipt of long-term dialysis or kidney transplant) over the 8-year period from January 1, 2000, through December 31, 2007, within each hospital referral region were identified using the USRDS’ Patients File based on zip code at the onset of ESRD. Denominator populations for each hospital referral region were identified using zip code data from the 2000 US Census. Using these 2 data sources, we calculated the crude average annual incidence of treated ESRD within each hospital referral region by age, race, and sex from 2000 through 2007.

The following measures of preparedness for ESRD were obtained from the USRDS’ Medical Evidence File: (1) receipt of care from a nephrologist before the onset of ESRD, (2) presence of optimal access for hemodialysis (ie, fistula vs graft or catheter), and (3) use of peritoneal dialysis (vs hemodialysis). Receipt of a kidney transplant as the initial treatment modality (preemptive transplant) was ascertained from the USRDS’ Patients File. Unlike hemodialysis, initial treatment of ESRD with either peritoneal dialysis or kidney transplant usually requires advance planning and is thus a marker of ESRD preparedness, with the caveat that not all patients will be eligible for or choose these modalities.

Information on end-of-life care practices was obtained from the USRDS’ Death File based on the report of the nephrologist (Centers for Medicare & Medicaid Services form 2746). Measures included whether dialysis was discontinued before death, whether the patient was receiving hospice care before death, and place of death (hospital vs home, dialysis unit, nursing home, or other).

Age at onset of ESRD, race, and sex were ascertained from the USRDS’ Patients File. The following characteristics were ascertained from the USRDS’ Medical Evidence File based on the report of the nephrologist at onset of ESRD (Centers for Medicare & Medicaid Services form 2728): (1) comorbid conditions, which included diabetes, coronary artery disease, peripheral arterial disease, stroke, congestive heart failure, chronic obstructive pulmonary disease, and cancer; (2) dual eligibility for Medicare and Medicaid; (3) clinical measures, which included each patient’s reported estimated glomerular filtration rate and body mass index (calculated as weight in kilograms divided by height in meters squared) at the onset of ESRD; (4) measures of functional status, which included whether patients needed assistance in activities of daily living, could ambulate, could transfer, and were living in a nursing home. The percentage of patients who died within 6 months of ESRD onset is also reported to provide a rough proxy measure for how sick patients were at the onset of ESRD, and is based on the date of death in the USRDS’ Patients File.

Patient characteristics, ESRD preparedness, and end-of-life care practices were described using means or proportions with 95% confidence intervals (CIs) and tests for trend across quintiles of the end-of-life expenditure index. For groups defined by age, race, and sex, the mean annual incidence of treated ESRD for hospital referral regions within each quintile of the end-of-life expenditure index was expressed as the number of cases of ESRD per 100 000 persons per year with 95% CIs. The ESRD incidence rate ratios with 95% CIs were calculated for the highest compared with the lowest quintile of the end-of-life expenditure index for groups defined by age, sex, and race.

From the lowest to the highest quintile of the end-of-life expenditure index, the proportion of black patients and the proportion of patients with dual coverage with Medicare and Medicaid increased substantially (Table 1). The prevalence of all comorbid conditions other than congestive heart failure decreased across end-of-life expenditure index quintiles, while that of congestive heart failure and all measures of functional impairment increased. Mortality within 6 months of ESRD onset also increased from the lowest to the highest quintile of the end-of-life expenditure index. Across end-of-life expenditure index quintiles, 6-month mortality ranged from 13.4% (95% CI, 11.6%–15.2%) to 15.2% (95% CI, 13.7%–16.7%) among those aged 65 through 69 years and from 30.3% (95% CI, 26.7%–33.8%) to 39.4% (95% CI, 36.8%–42.0%) among those aged 85 years or older (see eTable 2 at http://www.jama.com).

The incidence of treated ESRD was higher in blacks compared with whites and in men compared with women (Figure 1; see eTable 3 at http://www.jama.com). For all groups examined, incidence increased up to the ages of 80 through 85 years and then reached a plateau or decreased. For whites of all ages, the incidence of treated ESRD increased from the lowest to the highest quintile of the end-of-life expenditure index (Figure 1). Among blacks, there was no significant association between the incidence of treated ESRD and end-of-life expenditure index quintile in most age groups examined. However, an association was present in men aged 80 years or older and in women aged 85 years or older (Figure 1). For blacks and whites, the ESRD incidence rate ratio for the highest compared with the lowest quintile of the end-of-life expenditure index was greatest at older ages (Figure 2).

Patients living in the highest compared with the lowest quintile of the end-of-life expenditure index were less likely to have been under the care of a nephrologist before the onset of ESRD, less likely to initiate hemodialysis with a fistula (vs graft or catheter), less likely to select peritoneal dialysis (vs hemodialysis) as their initial modality, and less likely to receive a preemptive transplant (Table 2). These differences were still present after adjustment for baseline patient characteristics.

Existing literature indicates that rates of referral to hospice before death are extremely low among US patients receiving long-term dialysis,⁷ that nephrologists vary greatly in their approach to initiation and withdrawal of dialysis,⁸ and that most nephrologists have not be entrained and do not feel well prepared to address the end-of-life care needs of their patients.^9,10 Recognizing the need for uniform clinical practice guidelines in this area, a decade ago the Renal Physicians Association and the American Society of Nephrology published a clinical practice guideline on shared decision making in the appropriate initiation of and withdrawal from dialysis.¹¹ However, a recent study among US and Canadian nephrologists demonstrated that almost half of those surveyed were not aware of the guideline, suggesting significant under-use of this important resource.¹⁰ Similar to other forms of life support,^12,13 we observed a striking degree of regional variation in rates of dialysis discontinuation and other aspects of end-of-life care among older patients with ESRD. It seems unlikely that such variation could be explained by differences in patient prefer-encesalone—prior research indicates that preferences for end-of-life care among older Medicare beneficiaries do not vary greatly across regions.¹⁴ More likely, these findings highlight an urgent need for efforts to evaluate and optimize the quality of end-of-life care for older patients with ESRD to ensure provision of care that is congruent with patient preferences and values.

Referral to a nephrologist before the onset of ESRD should allow patients to receive information about prognosis and treatment choices and to engage in a process of shared decision making.¹¹ Almost one-third of patients in this cohort were not under the care of a nephrologist before the onset of ESRD, perhaps indicating that they had little opportunity to learn about their treatment options and prognosis before making a decision to initiate dialysis. Surprisingly, despite a higher density of nephrologists in higher intensity-of-care regions and evidence for greater reliance on specialist care in these areas,¹ rates of prior nephrology referral and other measures of preparedness for ESRD were less favorable compared with lower intensity-of-care regions. There are several possible explanations for this observation, although our study cannot provide definitive evidence to suggest which of these may be most important. Patients living in the higher intensity-of-care areas may be more likely to progress to ESRD unexpectedly (eg, as a result of acute kidney injury) or there may be other differences between patients living in higher intensity-of-care regions (which are predominantly metropolitan) and those living in lower end-of-life intensity-of-care regions (which are predominantly nonmetropolitan) that are not captured in our analyses (eg, knowledge, expectations, social support, and geographic isolation). Nephrologist practice style and environment (eg, level of competition and incentive structure) may also vary across regions, perhaps resulting in less emphasis on shared decision making in higher intensity-of-care regions. A better understanding of why a substantial number of older adults are not adequately prepared for dialysis—particularly in higher intensity-of-care regions in which dialysis treatment practices may be the most aggressive—may help to identify opportunities to improve the quality of advance care planning and shared decision making related to ESRD.

In contrast with end-of-life care practices and preparedness for ESRD, our data sources provide only indirect information about dialysis initiation practices. The USRDS registry does not include information on patients with advanced kidney disease who were not treated for ESRD with dialysis or kidney transplant. At the same time, the USCensus provides information on only a limited number of characteristics associated with risk of ESRD (ie, age, sex, and race). Thus, regional differences in the population incidence of ESRD reported herein may reflect differences in both treatment decisions and in unmeasured risk factors for progression to ESRD (eg, advanced kidney disease, diabetes, and hypertension).

Despite these limitations, we believe that, as demonstrated for other forms of life support,¹ a higher incidence of ESRD in higher intensity-of-care regions probably does reflect a greater willingness to initiate dialysis in these areas. Because the benefit of long-term dialysis is least certain in the oldest patients and in those with a higher burden of comorbidity,¹⁵ one might expect to find a greater degree of variability in initiation practices at older ages. Consistent with this possibility, regional differences in the incidence of ESRD did increase with age. Also suggestive of more liberal initiation practices in higher intensity-of-care regions, patients in these areas were more likely to die within 6 months of ESRD onset compared with those living in lower intensity-of-care regions, perhaps suggesting acceptance of sicker patients onto dialysis. Nevertheless, in the absence of a more suitable denominator population, regional differences in the incidence of ESRD presented herein are best viewed as suggestive rather than as definitive evidence of differences in practice.

There are several other limitations to this study. We did not have information on patient preferences so we could not rule out the possibility that observed differences in practice are explained by differences in patient preferences. In addition, USRDS data do not include detailed information on severity of comorbid conditions. However, it is unlikely that the large observed differences in practice described herein reflect residual confounding by differences in disease severity.