One- Emerging Issues: Early mortality on hemodialysis, & racial disparities in diabetic ESRD

Mortality and morbidity in the first year of ESRD therapy create a complex issue, as trends during small time periods may conflict with those shown during a period of twenty or more years. We have previously reported a period of 10–12 years in which there was little change in first-year mortality. In 2008, we began to explore first-year mortality by month, showing high rates after the first month, particularly in those age 65 and older, and noting higher estimated glomerular filtration rates (eGFRs), lower serum creatinines, and rising body mass indices (BMIs) at initiation. These factors appear to have competing effects on associated mortality, canceling each other out and thereby contributing to the appearance of no significant change in death rates over time. This year we look at more detailed assessments of cause-specific death and hospitalization to define the burden of illness in the early months on dialysis. In Figure 1.1, on the next page, we illustrate monthly first-year mortality rates. The pattern of overall mortality is striking, with a sharp peak in months two and three; mortality due to cardiovascular disease (CVD) and infection also peak in these months. Notable here is the increase since 1996 in mortality due to other causes. It is important to note that the Death Notification form (2746) changed in October, 2004; this may have led to increased reporting of other causes of death, a possibility which merits careful investigation. Patients frequently sign off dialysis in the first months because of complications such as infection and cardiovascular events, making difficult the classification of cause of death as well as the exact nature of the trends in these causes. Comparisons of just two different years, however, may provide insufficient perspective on these trends. Figures on the next spread, for example, illustrate the higher death rates in the early and mid-1980s; their subsequent decline may be secondary to major changes in the delivery of dialysis — the introduction of Kt/V to quantitate dialysis dose, the transition from acetate to bicarbonate dialysate, the use of more biocompatible dialyzer membranes, and the use of ultrafiltration control equipment to stabilize weight removal during dialysis. First- and second-month death rates for all causes then rose in the early 1990s, coincident with the rise in catheter placement rates (see Figure hp.13), and in contrast to the lower mortality rates consistently noted in the later months of the first year, a period when catheters are often converted to an internal access device, reducing infectious complications. While other hypotheses about these changing patterns should be explored, the infectious complications associated with catheters are well known and should be addressed, as more than 80 percent of dialysis patients still begin therapy with a catheter. Cause-specific death rates show a rise in the early 1990s, with cardiovascular and infectious rates beginning to fall in 2000 and 2003, respectively. The rising rates of death from other causes is a major concern needing to be explored in more detail, particularly as it relates to the changing categories on the Death Notification form. These shifting patterns should also be examined in terms of the primary causes of hospitalization, which underlie issues in mortality. We show that hospitalization rates in the first months of dialysis have changed little since 1991. There appear to be some recent downward trends, but it will take time to be sure of their exact direction. Rates of CVD hospitalizations appear to have changed little over the last 15 years, while infectious hospitalization rates have clearly increased — consistent with the rise in infectious mortality. Hospitalization due to other causes, in contrast, appears to have fallen slightly, and in a pattern markedly different from that of mortality due to other causes. With little recent change to the classifying of hospitalization cause, these comparisons provide insight into the potential impact of changes in death classification introduced by the latest Death Notification form. To more fully understand the potential impact of infectious events and complications on mortality in the first year of dialysis, we next investigate the use of outpatient intravenous antibiotics. Between 1998 and 2005, the number of prevalent patients receiving these medications rose from 30 to 44 percent, but fell to approximately 41 percent between 2006 and 2007 — consistent with the decline in inpatient vascular access infections (see Figure 6.4). The duration of time on antibiotics has increased since 1998, but plateaued in 2005. These data suggest that the change from non-cuffed to cuffed catheters, and the reduction in catheter placement rates, may reflect longer duration of catheter use and longer exposure to potential infections, manifested as a greater duration of antibiotic treatment. Those involved with vascular access use in dialysis patients need to address catheter-related complications by early conversion to an internal device, and more importantly, proactively address planned access placement in Stage 4 CKD patients to reduce the high rates of catheter use at dialysis initiation. We conclude with data on trends in the incidence of ESRD due to diabetes, with specific attention to racial and age disparities. In 2005 we reported that the incidence of diabetic ESRD had reached a plateau, and fallen in some older populations. In contrast to declines noted in the white population, however, rates were rising among younger African Americans and Native Americans. This year we show that these trends have continued. Factors which may contribute to the increasing rates of diabetic ESRD in the younger population may include the rising burden of obesity (see the NHANES data in Chapter One of Volume One), which has grown dramatically in younger African Americans — a major public health concern. These rising rates suggest that, to reverse these trends, public health programs are needed to address early detection and intervention related to diabetes and its renal complications. Rates of ESRD due to hypertension, the leading cause of ESRD in African Americans age 30–39, have plateaued, while rates due to glomerulonephritis have fallen (see Reference Tables A.2.2–3); changes in disease classification may be contributing to these findings.

Particularly among patients age 65 and older, all-cause mortality rates (adjusted for gender, race, and primary diagnosis) in the first year of hemodialysis peak in months two and three after the initiation of hemodialysis; mortality due to cardiovascular disease and infection is also greatest during this period.

The decline from the higher mortality rates seen in the early and mid-1980s may be explained by advances in the delivery of dialysis, including, as mentioned in the chapter introduction, the use of Kt/V to quantitate dialysis dose, the replacement of acetate with bicarbonate dialysate, the introduction of new dialyzer membranes, and the use of ultrafiltration control equipment to stabilize weight removal during the dialysis run. Across causes, first- and second-month mortality rates then rose during the 1990s, coinciding with the rise in catheter placement rates. Mortality rates for the later months of the year, in contrast, were lower; this is a period of therapy in which a catheter is often replaced with an internal access, and in which rates of infectious complications are therefore lower.

Notable in these data is the recent rise in mortality due to other causes. As seen in Figure 1.9, rates of hospitalization due to other causes have actually declined during the same period. There has been little recent change in the classifying of hospitalization cause; there have, however, been changes to the Death Notification form (2746), which may have led to increased classification of deaths as due to “other” causes. Mortality classified as withdrawal/failure to thrive, for instance, may be due in part to complications such as cardiovascular disease and infection. The effect of changes in the classification of causes of death will be investigated by the USRDS in future editions of the ADR. Figures 1.2, 1.3, 1.4, & 1.5; see page 364 for analytical methods. Incident hemodialysis patients age 20 & older.

Tables of mortality rates (adjusted for age, gender, race, and primary diagnosis) over the first months of dialysis further illustrate the changes over time seen in Figures 1.2–5. All-cause mortality in the second month of hemodialysis, for example, has remained above 450 deaths per 1,000 patient years at risk for the last decade, up from 377 in 1991. And while cardiovascular mortality rates in the second month have fallen from their peak of 238 in 1996, at 185 they remain higher than rates seen during the 1980s and early 1990s. Rates of mortality from infection are also greatest in the second month following ESRD initiation, at 50 deaths per 1,000 patient years — down from 65 in 2001, but 2.4 times greater than in 1981. Tables 1.a, 1.b, 1.c, & 1.d; see page 364 for analytical methods. Incident hemodialysis patients age 20 & older.

In the incident hemodialysis population, all-cause hospitalization rates (adjusted for gender, race, and primary diagnosis) have changed little across ages since 1991 and, in patients age 65 and older, are highest in the early months following initiation. For these older patients starting therapy in 2006, for example, rates varied from 4,350 admissions per 1,000 patient years in the first month of treatment to 1,956 during months 10–12, and rates across the first year fell 3.8–8.9 percent between 2005 and 2006; future data are needed to determine if this is the beginning of a true decline in these rates.

Hospitalization attributed to cardiovascular disease follows similar patterns, remaining relatively steady since the early 1990s, and highest, for patients age 65 and older, in the early months of ESRD. Among these patients, rates reach a high of 1,293 admissions per 1,000 patient years at risk in the first month and a low of 595 in months 10–12.

In contrast to all-cause and cardiovascular hospitalization, rates of hospitalization attributed to infection have climbed steeply since the early 1990s. Among patients age 45–64, for example, the rate in the fourth month of therapy was 619 admissions per 1,000 patient years in 2006, an 87 percent increase over 1991. In that same period, the rate during the first eight months grew 61–67 percent for patients age 65 and older.

The only steady decline in rates has occurred in admissions for other causes. Since 1991, among patients age 65 and older, these have fallen 12 percent for the first month of therapy, and 34–36 percent for the second and third months. This is in sharp contrast to the increase in reported mortality attributed to other causes, illustrated in Figure 1.5. The lack of any change in the reporting of causes of hospitalization lends support to the hypothesis that recent changes in the Death Notification form underlie the apparent increase in mortality due to other causes. Figures 1.6, 1.7, 1.8, & 1.9; see page 364 for analytical methods. Incident hemodialysis patients age 20 & older.

Among incident hemodialysis patients age 65 and older, hospital admission rates (adjusted for age, gender, race, and primary diagnosis) for all causes, cardiovascular disease, infection, and other causes are highest in the first months of therapy. The all-cause admission rate in 2006, for example, was 4,382 per 1,000 patient years at risk in the first month, compared to 2,856 in the third month and 1,959 in the last quarter of the year — decreases of 35 and 55 percent, respectively. Trends in cardiovascular admissions are similar, though rates are considerably lower. In 2006, the first-month rate was 1,325, falling to 847 in month three and 602 in the fourth quarter. The first-month rate of admissions for infection is nearly twice as high as that of the fourth quarter. Tables 1.e; see page 364 for analytical methods. Incident hemodialysis patients age 65 & older.

Between 1994 and 2007, the use of intravenous (IV) antibiotics in outpatient dialysis centers rose from 29.4 percent to over 41 percent, with vancomycin predominant. While its use declined from 1996 to 1998 — perhaps a reaction to initial reports of staphylococcus aureus with intermediate levels of vancomycin resistance — it rebounded in 1999; more than 27 percent of 2007 hemodialysis patients received vancomycin. The percentage of patients receiving cefazolin grew from 2.2 in 1994 to almost 18 in 2007, while since 2004 the use of broad-spectrum cephalosporins (ceftazidime, ceftriaxone, and cefotaxime) has remained stable at around 6 percent. Levofloxacin was used in 2.7 percent of hemodialysis patients in 2007. Use of other IV antibiotics is fairly limited, due primarily to the fact that the half-lives of most are insufficient to maintain therapeutic concentrations between dialysis sessions. Both antibiotic and culture use have declined slightly since 2005, which may reflect increased use of arteriovenous fistulas. Figures 1.10, 1.11; see page 365 for analytical methods. Point prevalent hemodialysis patients who survive the entire year.

Among patients with at least one outpatient culture during the year, antibiotics are used most often in those with at least one claim for both a blood culture and another type of culture, at 78 percent in 2006. This is an increase from 61 percent in 1994. Figure 1.12; see page 365 for analytical methods. Point prevalent hemodialysis patients who survive the entire year.

Antibiotics are often given empirically. In 2007, just 43 percent of patients receiving an IV outpatient antibiotic during the year had any claim for an outpatient culture. Of those patients given an outpatient IV antibiotic, 18 and 10 percent, respectively, had an outpatient culture claim in either the month of or the month prior to antibiotic use. Figures 1.13 & 1.14; see page 365 for analytical methods. Point prevalent hemodialysis patients who survive the entire year.

Between 1994 and 2005, the percentage of patients receiving no antibiotics during the year declined from 71 to 59, while the percentage with one or more months of outpatient therapy rose from 29.3 to 40.7. These trends started to reverse in 2005, which may reflect increased use of arteriovenous fistulas and declining use of arteriovenous grafts. Figure 1.15; see page 365 for analytical methods. Point prevalent hemodialysis patients who survive the entire year.

Paralleling the trends seen in Figure 1.15, the percentage of patients receiving more than one antibiotic in a month also increased between 1994 and 2005, from 8.9 to 16.9. This trend reversed in 2005, and 15.7 percent of patients in 2007 received more than one antibiotic in a month. Figure 1.16; see page 365 for analytical methods. Point prevalent hemodialysis patients who survive the entire year.

Over 92 percent of patients on combination antibiotics received vancomycin and an aminoglycoside in 1994; this fell to 56 percent in 2007. Use of cefazolin plus an aminoglycoside reached a peak of 13.8 percent in 2000–2001, declining to 8.1 percent in 2007. As use of cefazolin or vancomyin plus an aminoglycoside declined, that of vancomycin plus a broad-spectrum cephalosporin with gram negative coverage increased, as did that of other combinations. Interestingly, combined use of vancomycin and cefazolin has grown from 2.8 percent in 1994 to 12.7 in 2007. Figure 1.17; see page 365 for analytical methods. Point prevalent hemodialysis patients who survive the entire year.

This figure provides some explanation for how combination antibiotics are being used. In 2007, only 17 percent of patients had an outpatient claim for both vancomycin and cefazolin on the same day. The majority of combinations may, therefore, represent a switch from one antibiotic to another. Vancomycin is administered first 42 percent of the time, and cefazolin 41 percent. In contrast, when vancomycin or cefazolin is used in combination with an aminoglycoside, almost 80 percent of patients receive the combination on the same day. Figure 1.18; see page 365 for analytical methods. Point prevalent hemodialysis patients who survive all of 2007.

Body mass indices (BMIs), as illustrated by data from the CDC’s National Health and Nutrition Examination Surveys (NHANES), have been rising steadily since the late 1980s. A BMI of 30 kg/m2 or above was reported in 32.3 and 34.8 percent of male and female participants in 2003–2006, respectively, up from 19.5 and 24.7 percent in the 1988–1994 survey. Nearly one in three white female participants in the most recent survey had a BMI of 30 or above; among African American women, this number reached more than 53 percent — a dramatic increase from the 36.2 percent reported for 1988–1994. And 14.4 percent of African American women in the most recent survey had a BMI of 40 kg/m2 or above, a level nearly twice as high as that seen in the 1988–1994 survey, and considerably greater than the 6.4 percent reported for white female participants. Figure 1.19; see page 365 for analytical methods. NHANES participants age 20 & older.

Data from the CDC’s Behavioral Risk Factor Surveillance System (BRFSS) show that, between 1995–1997 and 2005–2007, the age-adjusted incidence of diagnosed diabetes increased across the country, with a range in growth from 51 percent in the West census region to 133 percent in the South. Among the states with available data, the greatest growth in reported diabetes is seen in Florida, Texas, and Idaho, each at 203–216 percent. Additional CDC data illustrate not only the growing percentage of people with diagnosed diabetes, but the marked differences by race. In 2006, 5.5 and 5.0 percent of white males and females, respectively, had a diagnosis of the disease; among African Americans, in contrast, the numbers reached 8.1 and 8.8 percent. Table 1.i & Figure 1.20; see page 365 for analytical methods. General U.S. population, age 18 & older (BRFSS data; 1.i); general U.S. population, adjusted for age (CDC data; 1.20).

Incident rates for ESRD caused by diabetes vary widely by age and race, both in absolute values and in rates of growth. Among whites age 30–39, for example, the incident rate (adjusted for gender) has fallen 7.6 percent since 2000, to 33.1 per million population. In African Americans and Native Americans of the same age, in contrast, the adjusted rate is now 122–123, 55 and 37 percent higher than in 2000. Different patterns are seen in older populations. Since 2000 the rate has fallen almost 12 percent for African Americans age 50–59, but has grown 2.4 percent for whites of the same age. Compared to rates in whites, the incidence of ESRD, however, still remains significantly greater among African Americans: 3.7 times higher among those age 30–39 and 40–49, 4.0 times higher for those 50–59, and 4.8 times higher for those age 20–29. Figure 1.21; see page 365 for analytical methods. Incident ESRD patients. See supplemental Reference Tables A.2.2–3 (on our website & CD-ROM) for data on the incidence of ESRD due to glomerulonephritis & hypertension.

Between 1990–1995 and 2002–2007, incident rates (adjusted for gender) of ESRD due to diabetes in whites age 30–39 remained unchanged in the upper quintile, averaging 139–142 per million population. Rates for African Americans, in contrast, increased 44.5 percent in the upper quintile, to 596.8 per million. Figure 1.22; see page 365 for analytical methods. Incident ESRD patients, age 30–39.