- Email: [email protected]
Improvements in dialysis patient mortality are associated with improvements in urea reduction ratio and hematocrit, 1999 to 2002
Improvements in Dialysis Patient Mortality Are Associated With Improvements in Urea Reduction Ratio and Hematocrit, 1999 to 2002 Robert A. Wolfe, PhD, Tempie E. Hulbert-Shearon, MS, Valarie B. Ashby, MA, Sangeetha Mahadevan, MS, and Friedrich K. Port, MD ● Background: Benefits in terms of reductions in mortality corresponding to improvements in Kidney Disease Outcomes Quality Initiative (K/DOQI) compliance for adequacy of dialysis dose and anemia control have not been documented in the literature. We studied changes in achieving K/DOQI guidelines at the facility level to determine whether those changes are associated with corresponding changes in mortality. Methods: Adjusted mortality and fractions of patients achieving K/DOQI guidelines for urea reduction ratios (URRs; >65%) and hematocrit levels (>33%) were computed for 2,858 dialysis facilities from 1999 to 2002 using national data for patients with end-stage renal disease. Linear and Poisson regression were used to study the relationship between K/DOQI compliance and mortality and between changes in compliance and changes in mortality. Results: In 2002, facilities in the lowest quintile of K/DOQI compliance for URR and hematocrit guidelines had 22% and 14% greater mortality rates (P < 0.0001) than facilities in the highest quintile, respectively. A 10-percentage point increase in fraction of patients with a URR of 65% or greater was associated with a 2.2% decrease in mortality (P ⴝ 0.0006), and a 10-percentage point increase in percentage of patients with a hematocrit of 33% or greater was associated with a 1.5% decrease in mortality (P ⴝ 0.003). Facilities in the highest tertiles of improvement for URR and hematocrit had a change in mortality rates that was 15% better than those observed for facilities in the lowest tertiles (P < 0.0001). Conclusion: Both current practice and changes in practices with regard to achieving anemia and dialysis-dose guidelines are associated significantly with mortality outcomes at the dialysis-facility level. Am J Kidney Dis 45:127–135. © 2004 by the National Kidney Foundation, Inc. INDEX WORDS: End-stage renal disease (ESRD); quality improvement; dose of dialysis; anemia management; standardized mortality ratio (SMR); hemodialysis (HD) survival.
A
FTER DECLINING between 1990 and 1995,1 adjusted mortality rates for US dialysis patients were relatively stable between 1995 and 2000.2 Substantial efforts are being made to improve the quality of end-stage renal disease (ESRD) care through the development of practice guidelines,3 reporting of clinical performance measures,4 continuing efforts by providers and ESRD Networks to improve quality, and the quality-assurance efforts of the certification process. Unfortunately, no magic bullet has yet been found that offers a simple way to reduce ESRD mortality rates. Whereas transplantation is associated with lower death rates than dialysis therapy,5 the potential for increasing transplantation rates is constrained by the availability of organs,6 rather than by the efforts of dialysis providers. The percentage of patients at or above Kidney Disease Outcomes Quality Initiative (K/DOQI) guidelines with regard to adequacy of hemodialysis treatment (perhaps measured best by Kt/V, but also measured by urea reduction ratio [URR]) and adequate anemia treatment (measured by hematocrit or hemoglobin levels) has been increasing in recent years.4 Many prior observational studies have shown an association of lower mortality rates with both
greater doses of dialysis (up to a URR of 75%)7,8 and greater hematocrit levels (up to a hematocrit of 33% to 36%),9 and there are plausible physiological mechanisms to explain such relationships. However, no national study has documented a reduction in mortality rates corresponding to these recent improvements in practice patterns. Two recent studies of the efficacy of greater doses of dialysis10,11 have shown little or no benefit of doses higher than K/DOQI guidelines. Published studies showing reduced mortality among patients with greater doses of dialysis and greater hematocrit levels have shown a relationship at the patient level of analysis, but these analyses
From the Department of Biostatistics, University of Michigan; and University Renal Research and Education Association, Ann Arbor, MI. Received March 5, 2004; accepted in revised form September 21, 2004. This study was supported through a grant from the Centers for Medicare and Medicaid Services (CMS contract no. 500-01-0056). Address reprint requests to Robert A. Wolfe, PhD, University of Michigan, KECC, 315 W Huron, Ste 240, Ann Arbor, MI 48103. E-mail: [email protected] © 2004 by the National Kidney Foundation, Inc. 0272-6386/04/4501-0014$30.00/0 doi:10.1053/j.ajkd.2004.09.023
American Journal of Kidney Diseases, Vol 45, No 1 (January), 2005: pp 127–135
127
128
WOLFE ET AL
are subject to bias because healthier patients are more likely to achieve a high dose of dialysis and high hematocrit levels. That is, healthy patients are more likely to reach targeted treatment goals because they are healthy, which is the reverse causation from the intended conclusion that achieving targeted goals leads to better health. This suggests that it is time to update the evidence that documents the relationship of mortality to the K/DOQI guidelines for dose of dialysis and hematocrit. To overcome some of the potential biases in patient-level analyses, we focus on dialysis-facility practices and standardized mortality at the dialysis-facility level. METHODS National data from the Centers for Medicare and Medicaid Services (CMS) were used for this study. These data were collected and summarized for all prevalent dialysis patients at 5,633 dialysis facilities in the United States during 1999 to 2002.
Measures Standardized mortality ratio. The standardized mortality ratio (SMR) compares the observed death rate for a group of patients with the death rate that would be expected based on national death rates for patients with the same characteristics.12 The SMR uses expected mortality calculated from an adjusted Cox model fit for all patients in the United States during the period of interest. We calculated expected deaths as ⫺ln(Si[ti]), where Si(t) is the survival curve from a Cox model adjusted to the characteristics of patient i, and ti is the amount of follow-up (patient-years at risk) for that patient during the year.13-15 SMR values for the United States were calculated to show the national trend in US mortality adjusted to a constant patient mix. For this calculation, a single adjusted Cox model was fit for all patient-years for patients on dialysis therapy during 1985 to 2002 (N ⫽ 3.2 million patient-years). This model was adjusted for age, race, sex, diabetes as cause of ESRD, and years of ESRD treatment. The resulting SMRs calculated for each year are relative to the average mortality for patients during the entire period. For all other analyses in this study, we calculated SMRs for each facility during 1999 to 2002. For this calculation, we fit a single adjusted Cox model for all patients on dialysis therapy during 1999 to 2002 only (N ⫽ 899,000 patientyears). This model was adjusted for additional factors not available for inclusion in national SMR results for 1985 to 2002. We adjusted this Cox model for age, race, sex, diabetes, years on dialysis therapy, facility comorbidity index, and facility average body mass index. The facility comorbidity index16 and average body mass index are based on all patients incident at the facility during 1999 to 2002 and are included at the facility, rather than patient, level because they are not available for all patients. The comorbidity index is based on comorbid conditions reported for
incident patients on the ESRD Medical Evidence Form (CMS 2728). We also controlled for age-adjusted population death rates by state and race based on the US population in 1998 to 2000.17 The additional adjustments make little difference in the resulting SMRs.18-20 Facility urea reduction ratio and hematocrit. For each facility, the percentage of patients with a URR of 65% or greater and a hematocrit value of 33% or greater was computed based on CMS Medicare dialysis claims (UB-92) submitted by the facility for each year for 1999 to 2002. The URR measure for each facility includes all hemodialysis patients with ESRD for at least 1 year and with at least 4 URR Medicare claims in the year from the facility. The hematocrit measure includes all dialysis patients with ESRD for at least 90 days and with at least 4 hematocrit Medicare claims in a year from the facility. For hematocrit calculation, patient claims starting before day 90 of ESRD and claims with hematocrit values less than 14% or greater than 60% were excluded. The average hematocrit value reported for each patient while treated by a particular facility during a year was calculated based on all claims during the year from the facility for the patient. Patients with fewer than 4 claims from a facility were excluded, and the percentage of remaining patients with an average hematocrit value of 33% or greater was calculated for each facility and year. Hematocrit values are required on dialysis bills that include erythropoietin treatment, but also were present on 16% of bills that did not include erythropoietin treatment. The hematocrit calculation included all patients with hematocrit values available regardless of erythropoietin treatment. For the URR calculation, patient claims starting before day 365 of ESRD for a patient and those with a missing URR category were excluded. The most common URR category reported (with ties going to the lower category) for each patient while treated by a particular facility during a year was calculated based on all claims during the year from the facility for the patient. Patients with fewer than 4 claims from a facility were excluded, and the percentage of remaining patients for whom the most common URR category was 65% or greater was calculated for each facility and year.
Statistical Analyses Exclusions. Veterans Affairs facilities, transplantationonly facilities, and facilities not known to be Medicare certified for each of the 4 years of this study were excluded from analyses (n ⫽ 1,087). Facilities with fewer than 5 patients included in hematocrit, URR, or SMR calculations or with less than 1 expected death for any year from 1999 to 2002 were excluded (n ⫽ 1,543). In addition, facilities treating less than 70% of patients with hemodialysis were excluded from these analyses (n ⫽ 145). After these exclusions, the final sample included 2,858 dialysis facilities, representing 709,000 patient-years (79% of all dialysis patient-years in mortality calculations before facility exclusions). Associations between practice patterns and mortality. Facilities were divided into quintiles according to the fraction of patients achieving K/DOQI guidelines for hematocrit in 1999. Facilities also were classified into 1 of these 5 groups according to hematocrit values for 2002. Poisson regression
ESRD MORTALITY AND FACILITY PRACTICE PATTERNS
129
Fig 1. Trend in SMRs for all dialysis patients, 1985 to 2002, adjusted for patient age, sex, race, diabetes as cause of ESRD, and years on dialysis therapy. Reference (SMR ⴝ 1.00) is average mortality during entire period.
was used to model the expected number of deaths based on the 10 resulting categories of hematocrit grouping (5 quintiles) and year (1999 and 2002). A similar analysis was performed for URR. Again, facilities were divided into quintiles according to the fraction of patients achieving K/DOQI guidelines for URR in 1999, and the same groupings were used to categorize facilities in 2002. Poisson regression was used to model the expected number of deaths based on the 10 resulting categories of URR (5 quintiles) and year (1999 and 2002). SMR was calculated as total observed deaths divided by total expected deaths for each group. Associations between changes in practice patterns and changes in mortality. Associations between changes in practice patterns and changes in mortality were analyzed at the facility level. Average change per year was calculated for each facility and for each of the 3 measures by means of linear regression of each measure (y) on year (x). The association of average change per year in SMR with the average change per year in percentage of patients with a URR of 65% or greater and percentage of patients with a hematocrit value of 33% or greater was analyzed by using linear regression models weighted by facility size. The models were checked for normality of residuals and outliers, and regression diagnostics were examined.
These trends were very similar when only patients in the 2,858 facilities in the averagechange analysis were included. Figures 2 and 3 show trends in URR and hematocrit during the 4-year period for patients treated at the 2,858 facilities in the averagechange analysis. The fraction of patients with a URR at or greater than K/DOQI guidelines (ⱖ65%) increased from 85% to 90% in the same period. The fraction of patients with a hematocrit value at or greater than K/DOQI guidelines (ⱖ33%) increased from 74% to 86% in the same period. Improvements in URR and hematocrit have occurred fairly steadily throughout the study period (⬃2%/y and 4%/y, respectively). These
RESULTS
Figure 1 shows trends in adjusted mortality between 1985 and 2002 for all dialysis patients in the United States. Mortality for dialysis patients in the United States was 5.4% lower in 2002 than 1999. The decrease from 1999 to 2000 was 3.3%, and from 2001 to 2002, was 2.3%.
Fig 2. Trend in percentage of patients receiving an adequate URR, 1999 to 2002. N ⴝ 2,858 facilities.
130
Fig 3. Trend in percentage of patients receiving adequate anemia treatment, 1999 to 2002. N ⴝ 2,858 facilities.
trends were very similar when all patients were included regardless of facility type and size. In both 1999 and 2002, the SMRs were lower for facilities achieving greater levels of K/DOQI compliance with regard to URR (Fig 4) and hematocrit (Fig 5). For each figure, categories were chosen so that 20% of facilities were in each category in 1999 (quintiles). For hematocrit, the distribution of facilities shifted since 1999; thus, in 2002, there were 55% of facilities in the highest category of hematocrit and only 3% in the lowest category. There also was a shift
WOLFE ET AL
in URR since 1999, with 33% of facilities in the highest category of URR and only 7% in the lowest category in 2002. Although the number of facilities in the lower categories was much smaller in 2002 because of overall improvements in adequacy of dialysis and anemia treatment, mortality associated with similar levels of compliance was similar in 1999 and 2002. In 2002, facilities with more than 95% of patients achieving the K/DOQI guidelines for URR had 22% lower mortality (relative risk, 0.78; P ⬍ 0.0001) than those with less than 78% of patients achieving the K/DOQI guidelines (last versus first bars). Similarly, facilities with more than 87% of patients achieving the K/DOQI guidelines for hematocrit had 14% lower mortality (relative risk, 0.86; P ⬍ 0.0001) than those with less than 64% of patients achieving the K/DOQI guidelines (last versus first bars). Table 1 lists 9 groups of facilities formed by simultaneous classification with regard to tertile of improvement (average change between 1999 and 2002) with respect to K/DOQI compliance for hematocrit and URR. (Tertiles are labeled in Table 1 by tertile range and “small”/“none or decline,” “average,” or “large” improvement.)
Fig 4. SMRs for facilities by percentage of patients with a URR of 65% or greater, 1999 and 2002. Categories are defined as approximate quintiles for 1999. The number of facilities in each group is shown at the bottom of each bar. Reference group is facilities with 90.6% to 95.1% of patients with a URR of 65% or greater in 1999. The relative mortality risk for facilities in the highest quintile of percentage of patients with a URR of 65% or greater compared with the lowest quintile is 0.93/1.19 ⴝ 0.78 (P < 0.0001) for 2002. N ⴝ 2,858 facilities.
ESRD MORTALITY AND FACILITY PRACTICE PATTERNS
131
Fig 5. SMRs for facilities by percentage of patients with a hematocrit of 33% or greater, 1999 and 2002. Categories are defined as approximate quintiles for 1999. The number of facilities in each group is shown at the bottom of each bar. Reference group is facilities with 81.2% to 87.1% of patients with a hematocrit value of 33% or greater in 1999. The relative mortality risk for facilities in the highest quintile of percentage of patients with a hematocrit value of 33% or greater compared with the lowest quintile is 0.94/1.09 ⴝ 0.86 (P < 0.0001) for 2002. N ⴝ 2,858 facilities.
For hematocrit, the fraction of patients achieving K/DOQI guidelines increased by more that 20.6 percentage points at one third of facilities, between 6.0 and 20.6 percentage points for the middle third of facilities, and less than 6.0 percentage points for the final third of facilities. The analogous tertiles for increases in the fraction of patients achieving K/DOQI guidelines with respect to URR were more than 9.4 percentage points, between 0.2 and 9.4 percentage points, and less than 0.2 percentage points. Within the resulting 9 categories of facilities, Table 1 lists average changes in standardized mortality between 1999 and 2002. Improvement in mortality generally is successively greater moving to the right within each row and moving down within each column. This suggests that at each level of improvement in hematocrit (within each row), mortality improvement was greater as URR improvement was greater. Similarly, for each level of URR improvement (within a column), there was more improvement in mortality as improvement in hematocrit was greater. The 359 facilities with the least improvements for URR and hematocrit (upper left cell) experienced a 3% increase in mortality, whereas the 376 facilities
with the most improvements for URR and hematocrit (lower right cell) experienced a 12.4% decline in mortality during this period, on average. Multiple regression analysis shows that compliance with the K/DOQI guidelines for URR and hematocrit have independent effects on mortality. A 10-percentage point increase in fraction of patients with a URR of 65% or greater was associated with a 2.2% decrease in mortality (P ⫽ 0.0006), and a 10-percentage point increase in percentage of patients with a hematocrit value of 33% or greater was associated with a 1.5% decrease in mortality (P ⫽ 0.003) according to regression coefficient estimates. There was no significant interaction between effects of URR and hematocrit slopes on SMR slope (P ⫽ 0.75). Regression diagnostics indicated that the normality assumption was reasonably satisfied, and estimates were not overly influenced by individual data points. DISCUSSION
These results corroborate other studies that have shown increases in hematocrit and dose of dialysis levels for patients with ESRD treated
132
WOLFE ET AL Table 1. Average Percentage of Improvement in Mortality During 1999 to 2002 by Level of Improvement in Practice Patterns Improvement in Percent of Patients With URR ⱖ65%* (no. of facilities)
Improvement in Percent of Patients With Hematocrit ⱖ33%*
None or Decline (⬍0.2%)
Average (0.2%-9.4%)
Large (⬎9.4%)
Small (⬍6.0%) Average (6.0%-20.6%) Large (⬎20.6%)
⫺3% (359) 1.9% (323) 1.2% (262)
1.6% (328) 6.9% (336) 9.7% (306)
2% (257) 12.6% (311) 12.4% (376)
NOTE. Number of facilities (tertiles of hematocrit and tertiles of URR) in each category is shown in parentheses. N ⫽ 2,858 facilities, weighted by average number of patients in the statistic. Total change during 4 years as a percentage of deaths. Positive percentages correspond to lower mortality. *Total percentage point change over 4 years.
with hemodialysis in the United States in recent years. Improvements persisted through the end of 2002. Although there is some variation among types of facility and regions of the country, which are not shown in this report, the improvement largely has been broad based, and the vast majority of facilities have improved K/DOQI compliance for both URR and hematocrit in recent years (Table 1). The national reduction in mortality reported here is encouraging because mortality for patients with ESRD has remained relatively flat during the later years of the 1990s. Although the magnitude of this reduction in mortality is modest, ie, 5.7% between 1999 and 2002, meaning 19 instead of 20 deaths in a typical hemodialysis facility, it represents a reduction of 3,000 deaths/y among more than 60,000 deaths2 annually for US hemodialysis patients. This national reduction in mortality undoubtedly is caused by a wide variety of factors. However, at the facility level, mortality is associated with improvements in both dialysis dose and hematocrit; thus, it is likely that some of this reduction at the national level is associated with improvements in practice patterns. Because this is an observational study, one cannot infer a causal relationship at either the facility level or nationally. Many previous studies documented that mortality is lower among patients receiving higher dialysis doses and those achieving greater hematocrit levels.7-10,21,22 As the norms of standard practice have moved toward higher doses of dialysis, it has become possible to study this relationship at higher doses, and it now is apparent that whereas mortality rates are lower among patients receiving a greater URR, the relation-
ship is flatter at doses greater than a URR of 65% or greater. Port et al21 reported no difference in mortality among males for a URR of 70% or greater, but continued benefits for greater doses among females. These results are consistent with the recent Hemodialysis Study.10 The results here support these prior results and suggest that differences among facilities in achieving K/DOQI guidelines for URR levels are associated significantly with corresponding differences in mortality (Fig 4). Differences in achieved URRs (⬍78% versus ⬎95% at K/DOQI guidelines) and mortality risks (SMR of 1.20 versus 0.92 for 2002) between the worst and best quintiles of facilities are both substantial. These findings do not disagree with the Hemodialysis Study because the Hemodialysis Study did not evaluate patients receiving a URR less than 65%.10 Differences in the fraction of patients achieving a hematocrit value of 33% or greater at the facility level (⬍62% versus ⬎87%) and the corresponding difference in mortality risk (SMR of 1.15 versus 0.94 for 2002) between the worst and best quintiles of facilities also are substantial. These analyses have been adjusted for many patient characteristics; it therefore is unlikely that the relationship shown is caused by unmeasured differences in patient comorbidity. This suggests that K/DOQI goals can be achieved for nearly all patients at most facilities, and the benefits of doing so include longer lifetimes for patients at those facilities. It is possible that changes in URR or hematocrit are, in part, a result of changes in catheter use. During 1999 to 2002, the use of temporary catheters has been decreasing, although there has been increased use of long-term catheters, lead-
ESRD MORTALITY AND FACILITY PRACTICE PATTERNS
ing to an overall increase in catheter use.23 Although dialysis dose and anemia management are unlikely to be independent of other clinical factors, such as catheter use, this study finds an independent association between each of these factors and mortality at the facility level. Facilities that improved dialysis dose, but not hematocrit, for patients still had improvements in mortality, and vice versa. Many facilities improved on only 1 of these 2 factors, not both. Prior facility-based studies have shown that differences in facility practice patterns are associated with differences in mortality.24,25 This is one of the first studies that shows that changes in practice patterns at a facility over time are associated with concurrent changes in mortality. The genesis of the substantial improvement in achieving K/DOQI targets is unclear, but could include the effects of continuous quality improvement efforts by facilities, chains, networks, or the CMS. There have been few studies that document the effect of such changes in anemia and URR practices on such patient outcomes as mortality. As listed in Table 1, facilities that achieved greater improvements in K/DOQI practices also achieved greater reductions in patient mortality. Limitations of this study are as follows. These analyses do not look at patient-by-patient improvement in URR, hematocrit, or mortality, but instead look at aggregates and associations between these aggregates. The associations may not hold at the patient level. We used national Medicare data to study the relationship between practice patterns and mortality at the facility level in 2 ways. First, we studied the relationship of standardized mortality with fractions of patients currently achieving K/DOQI guidelines with regard to URR and hematocrit levels at each facility. Second, we studied the relationship between changes in K/DOQI compliance and changes in mortality using data from 1999 to 2002 to determine whether facilities that improved their K/DOQI compliance levels also had a corresponding decrease in mortality. These analyses were performed at the facility level, which avoids much of the potential bias caused by the healthypatient effect, ie, that patients with greater URR or hematocrit levels may be healthier patients to begin with. The level of achieving practice guidelines was measured at the facility level (as the
133
percentage of patients achieving the guideline), and this measure was used to predict mortality for all patients at that facility. The facility-level measure achieving practice guidelines can be interpreted as a facility-level “intent-to-treat protocol” for all patients at that facility. Facilitylevel analyses do not completely avoid the healthy-patient bias because facilities with a more difficult patient mix are likely to have greater mortality and lower compliance with K/DOQI guidelines because of that patient mix. However, facility-level analyses can reduce this potential bias dramatically. The analysis of changes in this report further reduces this potential for bias because each facility acts as its own control in the calculation of change per year. The facility URR and hematocrit measures studied here were for Medicare patients only, whereas facility mortality was measured for all patients, both Medicare and non-Medicare, at the facilities. This may weaken the associations between mortality and URR and hematocrit, although it is likely that values for Medicare patients at a facility reflect practices for all patients at the facility. One limitation to this study is that patient URR and hematocrit measures are self-reported by facilities. In addition, URR is reported in categories, rather than as a continuous variable. A validation study by Frankenfield et al26 showed that these claims data are a reliable source of information, particularly when classifying patients as above or below threshold values. URR and anemia management may not be independent of other clinical practices in fact or in a statistical sense. Those other practices may (likely do) influence mortality as much as or more than management of these 2 clinical practice parameters. Examples may be the correlation of hemoglobin and albumin levels27 or the use of catheters.28 Unfortunately, information for serum albumin level and vascular access is not available longitudinally in the CMS data. The present study focuses on 2 practices at the dialysis-unit level that may be modified more easily than serum albumin concentration. All observational studies, including this one, are subject to potential bias resulting from unmeasured differences in patient mix that affect both practice measures and patient outcomes. It is plausible that some component of the relation-
134
WOLFE ET AL
ships shown here is caused by unmeasured differences in patient mix among facilities. For example, it might be more difficult to achieve high URR and hematocrit levels at facilities with sicker patients, and those facilities also would be expected to have greater mortality rates. Patient mix at a facility is less likely to change substantially during a 4-year period; therefore, correlations among change scores, as listed in Table 1, are less likely to be affected by such potential bias. Additionally, analyses reported here are adjusted for many patient characteristics, including age and diabetes, which makes it less likely that relationships reported here are caused by unmeasured patient characteristics. In conclusion, the wide variation among facilities in achieving K/DOQI guidelines and that these analyses are adjusted for many patient factors suggests that differences in achievement of K/DOQI guidelines might be caused, in part, by practice pattern differences among facilities, rather than differences in patient mix. The observation that greater improvements in practice patterns, measured by increasing adherence to URR and hematocrit guidelines, correlate with greater improvements in mortality at the facility level strongly suggests benefits from greater adherence to dialysis dose and anemia guidelines. REFERENCES 1. US Renal Data System: USRDS 1999 Annual Data Report. The National Institutes of Health, National Institute of Diabetes and Digestive and Kidney Diseases, Bethesda, MD, 1999 2. US Renal Data System: USRDS 2003 Annual Data Report. The National Institutes of Health, National Institute of Diabetes and Digestive and Kidney Diseases, Bethesda, MD, 2003 3. National Kidney Foundation: K/DOQI Clinical Practice Guidelines for Nutrition in Chronic Renal Failure. Am J Kidney Dis 35:S1-S140, 2000 (suppl 2) 4. Centers for Medicare and Medicaid Services: 2002 Annual Report, End-Stage Renal Disease Clinical Performance Measures Project. Baltimore, MD, Department of Health and Human Services, Centers for Medicare and Medicaid Services, Center for Beneficiary Choices, 2002 5. Wolfe RA, Ashby VB, Milford EL, et al: Comparison of mortality in all patients on dialysis, patients on dialysis awaiting transplantation, and recipients of a first cadaveric transplant. N Engl J Med 341:1725-1730, 1999 6. US Organ Procurement and Transplantation Network and the Scientific Registry for Transplant Recipients: 2002 Annual Report of the US Organ Procurement and Transplantation Network and the Scientific Registry for Transplant
Recipients: Transplant Data 1992-2001: Department of Health and Human Services, Health Resources and Services Administration, Office of Special Programs, Division of Transplantation, Rockville, MD; United Network for Organ Sharing, Richmond, VA; University Renal Research and Education Association, Ann Arbor, MI, 2002 7. Wolfe RA, Ashby VB, Daugirdas JT, Agodoa LY, Jones CA, Port FK: Body size, dose of hemodialysis, and mortality. Am J Kidney Dis 35:80-88, 2000 8. Port FK, Ashby VB, Dhingra RK, Roys E, Wolfe RA: Dialysis dose and body mass index are strongly associated with survival in hemodialysis patients. J Am Soc Nephrol 13:1061-1066, 2002 9. Collins AJ, Li S, St Peter W, et al: Death, hospitalization, and economic associations among incident hemodialysis patients with hematocrit values of 36 to 39%. J Am Soc Nephrol 12:2465-2473, 2001 10. Eknoyan G, Beck GJ, Cheung AK, et al, the Hemodialysis (HEMO) Study Group: Effect of dialysis dose and membrane flux in maintenance hemodialysis. N Engl J Med 347:2010-2019, 2002 11. Paniagua R, Amato D, Vonesh E, et al: Effects of increased peritoneal clearances on mortality rates in peritoneal dialysis: ADEMEX, a prospective, randomized, controlled trial. J Am Soc Nephrol 13:1307-1320, 2002 12. Wolfe RA, Gaylin DS, Port FK, Held PJ, Wood CL: Using USRDS generated mortality tables to compare local ESRD mortality rates to national rates. Kidney Int 42:991996, 1992 13. SAS Institute Inc: SAS/STAT User’s Guide, version 8. Cary, NC, SAS Institute Inc, 1999, pp 2598-2599 14. Andersen PK, Borgun O, Gill RD, Keiding N: Statistical Models Based on Counting Processes. New York, NY, Springer-Verlag, 1993, pp 334, 406-407 15. Collett D: Modeling Survival Data in Medical Research. London, UK, Chapman & Hall, 1994, p 153, equation 5.6, p 151, equation 5.1 16. Wolfe RA, Ashby VB, Port FK: 1993 DMMS comorbidity index validated by Medical Evidence Form data. J Am Soc Nephrol 11:1300A, 2000 (abstr) 17. National Center for Health Statistics. Health, United States, 2002 With Rural and Urban Health Chartbook. Hyattsville, MD, Centers for Disease Control and Prevention, Health and Human Services Department, 2002, pp 117-118 18. Wolfe RA, Ashby VB, Hulbert-Shearon TE, Port FK: New dialysis facility mortality statistics (SMRs) adjust for more patient characteristics. J Am Soc Nephrol 12:1802A, 2001 (abstr) 19. Wolfe RA, Ashby VB, Port FK, et al: The association between regional death rates and standardized mortality ratios (SMR) by health service areas (HSA) and race and sex group. J Am Soc Nephrol 9:230A, 1998 (abstr) 20. Ashby VB, Wolfe RA, Loos ME, Port FK: The effect of comorbidities on facility standardized mortality ratios. J Am Soc Nephrol 9:197A, 1998 (abstr) 21. Port FK, Wolfe RA, Hulbert-Shearon TE, McCullough KP, Ashby VB, Held PJ: High dialysis dose is associated with lower mortality among females but not among males. Am J Kidney Dis 43:1014-1023, 2004
ESRD MORTALITY AND FACILITY PRACTICE PATTERNS
22. Ma J, Ebben J, Xia H, Collins A: Hematocrit level and associated mortality in hemodialysis patients. J Am Soc Nephrol 10:610-619, 1999 23. Centers for Medicare and Medicaid Services: 2002 Annual Report, End-Stage Renal Disease Clinical Performance Measures Project. Baltimore, MD, Department of Health and Human Services, Centers for Medicare and Medicaid Services, Center for Beneficiary Choices, 2003 24. Lowrie EG, Teng M, Lacson E, Lew N, Lazarus JM, Owen WF: Association between prevalent care process measures and facility specific mortality rates. Kidney Int 60:1917-1929, 2001 25. McClellan WM, Soucie JM, Flanders WD: Mortality in end-stage renal disease is associated with facility-to-
135
facility differences in adequacy of hemodialysis. J Am Soc Nephrol 9:1940-1947, 1998 26. Frankenfield DL, Brier ME, Bedinger MR, et al: Comparison of urea reduction ratio and hematocrit data reported in different data systems: Results from the Centers for Medicare and Medicaid Services and the Renal Network Inc. Am J Kidney Dis 41:433-441, 2003 27. Pisoni RL, Bragg-Gresham JL, Young EW, et al: Anemia management outcomes from 12 countries in the Dialysis Outcomes and Practice Patterns Study (DOPPS). Am J Kidney Dis 44:94-111, 2004 28. Dhingra RK, Young EW, Hulbert-Shearon TE, Leavey SF, Port FK: Type of vascular access and mortality in US hemodialysis patients. Kidney Int 60:1443-1451, 2001
A
FTER DECLINING between 1990 and 1995,1 adjusted mortality rates for US dialysis patients were relatively stable between 1995 and 2000.2 Substantial efforts are being made to improve the quality of end-stage renal disease (ESRD) care through the development of practice guidelines,3 reporting of clinical performance measures,4 continuing efforts by providers and ESRD Networks to improve quality, and the quality-assurance efforts of the certification process. Unfortunately, no magic bullet has yet been found that offers a simple way to reduce ESRD mortality rates. Whereas transplantation is associated with lower death rates than dialysis therapy,5 the potential for increasing transplantation rates is constrained by the availability of organs,6 rather than by the efforts of dialysis providers. The percentage of patients at or above Kidney Disease Outcomes Quality Initiative (K/DOQI) guidelines with regard to adequacy of hemodialysis treatment (perhaps measured best by Kt/V, but also measured by urea reduction ratio [URR]) and adequate anemia treatment (measured by hematocrit or hemoglobin levels) has been increasing in recent years.4 Many prior observational studies have shown an association of lower mortality rates with both
greater doses of dialysis (up to a URR of 75%)7,8 and greater hematocrit levels (up to a hematocrit of 33% to 36%),9 and there are plausible physiological mechanisms to explain such relationships. However, no national study has documented a reduction in mortality rates corresponding to these recent improvements in practice patterns. Two recent studies of the efficacy of greater doses of dialysis10,11 have shown little or no benefit of doses higher than K/DOQI guidelines. Published studies showing reduced mortality among patients with greater doses of dialysis and greater hematocrit levels have shown a relationship at the patient level of analysis, but these analyses
From the Department of Biostatistics, University of Michigan; and University Renal Research and Education Association, Ann Arbor, MI. Received March 5, 2004; accepted in revised form September 21, 2004. This study was supported through a grant from the Centers for Medicare and Medicaid Services (CMS contract no. 500-01-0056). Address reprint requests to Robert A. Wolfe, PhD, University of Michigan, KECC, 315 W Huron, Ste 240, Ann Arbor, MI 48103. E-mail: [email protected] © 2004 by the National Kidney Foundation, Inc. 0272-6386/04/4501-0014$30.00/0 doi:10.1053/j.ajkd.2004.09.023
American Journal of Kidney Diseases, Vol 45, No 1 (January), 2005: pp 127–135
127
128
WOLFE ET AL
are subject to bias because healthier patients are more likely to achieve a high dose of dialysis and high hematocrit levels. That is, healthy patients are more likely to reach targeted treatment goals because they are healthy, which is the reverse causation from the intended conclusion that achieving targeted goals leads to better health. This suggests that it is time to update the evidence that documents the relationship of mortality to the K/DOQI guidelines for dose of dialysis and hematocrit. To overcome some of the potential biases in patient-level analyses, we focus on dialysis-facility practices and standardized mortality at the dialysis-facility level. METHODS National data from the Centers for Medicare and Medicaid Services (CMS) were used for this study. These data were collected and summarized for all prevalent dialysis patients at 5,633 dialysis facilities in the United States during 1999 to 2002.
Measures Standardized mortality ratio. The standardized mortality ratio (SMR) compares the observed death rate for a group of patients with the death rate that would be expected based on national death rates for patients with the same characteristics.12 The SMR uses expected mortality calculated from an adjusted Cox model fit for all patients in the United States during the period of interest. We calculated expected deaths as ⫺ln(Si[ti]), where Si(t) is the survival curve from a Cox model adjusted to the characteristics of patient i, and ti is the amount of follow-up (patient-years at risk) for that patient during the year.13-15 SMR values for the United States were calculated to show the national trend in US mortality adjusted to a constant patient mix. For this calculation, a single adjusted Cox model was fit for all patient-years for patients on dialysis therapy during 1985 to 2002 (N ⫽ 3.2 million patient-years). This model was adjusted for age, race, sex, diabetes as cause of ESRD, and years of ESRD treatment. The resulting SMRs calculated for each year are relative to the average mortality for patients during the entire period. For all other analyses in this study, we calculated SMRs for each facility during 1999 to 2002. For this calculation, we fit a single adjusted Cox model for all patients on dialysis therapy during 1999 to 2002 only (N ⫽ 899,000 patientyears). This model was adjusted for additional factors not available for inclusion in national SMR results for 1985 to 2002. We adjusted this Cox model for age, race, sex, diabetes, years on dialysis therapy, facility comorbidity index, and facility average body mass index. The facility comorbidity index16 and average body mass index are based on all patients incident at the facility during 1999 to 2002 and are included at the facility, rather than patient, level because they are not available for all patients. The comorbidity index is based on comorbid conditions reported for
incident patients on the ESRD Medical Evidence Form (CMS 2728). We also controlled for age-adjusted population death rates by state and race based on the US population in 1998 to 2000.17 The additional adjustments make little difference in the resulting SMRs.18-20 Facility urea reduction ratio and hematocrit. For each facility, the percentage of patients with a URR of 65% or greater and a hematocrit value of 33% or greater was computed based on CMS Medicare dialysis claims (UB-92) submitted by the facility for each year for 1999 to 2002. The URR measure for each facility includes all hemodialysis patients with ESRD for at least 1 year and with at least 4 URR Medicare claims in the year from the facility. The hematocrit measure includes all dialysis patients with ESRD for at least 90 days and with at least 4 hematocrit Medicare claims in a year from the facility. For hematocrit calculation, patient claims starting before day 90 of ESRD and claims with hematocrit values less than 14% or greater than 60% were excluded. The average hematocrit value reported for each patient while treated by a particular facility during a year was calculated based on all claims during the year from the facility for the patient. Patients with fewer than 4 claims from a facility were excluded, and the percentage of remaining patients with an average hematocrit value of 33% or greater was calculated for each facility and year. Hematocrit values are required on dialysis bills that include erythropoietin treatment, but also were present on 16% of bills that did not include erythropoietin treatment. The hematocrit calculation included all patients with hematocrit values available regardless of erythropoietin treatment. For the URR calculation, patient claims starting before day 365 of ESRD for a patient and those with a missing URR category were excluded. The most common URR category reported (with ties going to the lower category) for each patient while treated by a particular facility during a year was calculated based on all claims during the year from the facility for the patient. Patients with fewer than 4 claims from a facility were excluded, and the percentage of remaining patients for whom the most common URR category was 65% or greater was calculated for each facility and year.
Statistical Analyses Exclusions. Veterans Affairs facilities, transplantationonly facilities, and facilities not known to be Medicare certified for each of the 4 years of this study were excluded from analyses (n ⫽ 1,087). Facilities with fewer than 5 patients included in hematocrit, URR, or SMR calculations or with less than 1 expected death for any year from 1999 to 2002 were excluded (n ⫽ 1,543). In addition, facilities treating less than 70% of patients with hemodialysis were excluded from these analyses (n ⫽ 145). After these exclusions, the final sample included 2,858 dialysis facilities, representing 709,000 patient-years (79% of all dialysis patient-years in mortality calculations before facility exclusions). Associations between practice patterns and mortality. Facilities were divided into quintiles according to the fraction of patients achieving K/DOQI guidelines for hematocrit in 1999. Facilities also were classified into 1 of these 5 groups according to hematocrit values for 2002. Poisson regression
ESRD MORTALITY AND FACILITY PRACTICE PATTERNS
129
Fig 1. Trend in SMRs for all dialysis patients, 1985 to 2002, adjusted for patient age, sex, race, diabetes as cause of ESRD, and years on dialysis therapy. Reference (SMR ⴝ 1.00) is average mortality during entire period.
was used to model the expected number of deaths based on the 10 resulting categories of hematocrit grouping (5 quintiles) and year (1999 and 2002). A similar analysis was performed for URR. Again, facilities were divided into quintiles according to the fraction of patients achieving K/DOQI guidelines for URR in 1999, and the same groupings were used to categorize facilities in 2002. Poisson regression was used to model the expected number of deaths based on the 10 resulting categories of URR (5 quintiles) and year (1999 and 2002). SMR was calculated as total observed deaths divided by total expected deaths for each group. Associations between changes in practice patterns and changes in mortality. Associations between changes in practice patterns and changes in mortality were analyzed at the facility level. Average change per year was calculated for each facility and for each of the 3 measures by means of linear regression of each measure (y) on year (x). The association of average change per year in SMR with the average change per year in percentage of patients with a URR of 65% or greater and percentage of patients with a hematocrit value of 33% or greater was analyzed by using linear regression models weighted by facility size. The models were checked for normality of residuals and outliers, and regression diagnostics were examined.
These trends were very similar when only patients in the 2,858 facilities in the averagechange analysis were included. Figures 2 and 3 show trends in URR and hematocrit during the 4-year period for patients treated at the 2,858 facilities in the averagechange analysis. The fraction of patients with a URR at or greater than K/DOQI guidelines (ⱖ65%) increased from 85% to 90% in the same period. The fraction of patients with a hematocrit value at or greater than K/DOQI guidelines (ⱖ33%) increased from 74% to 86% in the same period. Improvements in URR and hematocrit have occurred fairly steadily throughout the study period (⬃2%/y and 4%/y, respectively). These
RESULTS
Figure 1 shows trends in adjusted mortality between 1985 and 2002 for all dialysis patients in the United States. Mortality for dialysis patients in the United States was 5.4% lower in 2002 than 1999. The decrease from 1999 to 2000 was 3.3%, and from 2001 to 2002, was 2.3%.
Fig 2. Trend in percentage of patients receiving an adequate URR, 1999 to 2002. N ⴝ 2,858 facilities.
130
Fig 3. Trend in percentage of patients receiving adequate anemia treatment, 1999 to 2002. N ⴝ 2,858 facilities.
trends were very similar when all patients were included regardless of facility type and size. In both 1999 and 2002, the SMRs were lower for facilities achieving greater levels of K/DOQI compliance with regard to URR (Fig 4) and hematocrit (Fig 5). For each figure, categories were chosen so that 20% of facilities were in each category in 1999 (quintiles). For hematocrit, the distribution of facilities shifted since 1999; thus, in 2002, there were 55% of facilities in the highest category of hematocrit and only 3% in the lowest category. There also was a shift
WOLFE ET AL
in URR since 1999, with 33% of facilities in the highest category of URR and only 7% in the lowest category in 2002. Although the number of facilities in the lower categories was much smaller in 2002 because of overall improvements in adequacy of dialysis and anemia treatment, mortality associated with similar levels of compliance was similar in 1999 and 2002. In 2002, facilities with more than 95% of patients achieving the K/DOQI guidelines for URR had 22% lower mortality (relative risk, 0.78; P ⬍ 0.0001) than those with less than 78% of patients achieving the K/DOQI guidelines (last versus first bars). Similarly, facilities with more than 87% of patients achieving the K/DOQI guidelines for hematocrit had 14% lower mortality (relative risk, 0.86; P ⬍ 0.0001) than those with less than 64% of patients achieving the K/DOQI guidelines (last versus first bars). Table 1 lists 9 groups of facilities formed by simultaneous classification with regard to tertile of improvement (average change between 1999 and 2002) with respect to K/DOQI compliance for hematocrit and URR. (Tertiles are labeled in Table 1 by tertile range and “small”/“none or decline,” “average,” or “large” improvement.)
Fig 4. SMRs for facilities by percentage of patients with a URR of 65% or greater, 1999 and 2002. Categories are defined as approximate quintiles for 1999. The number of facilities in each group is shown at the bottom of each bar. Reference group is facilities with 90.6% to 95.1% of patients with a URR of 65% or greater in 1999. The relative mortality risk for facilities in the highest quintile of percentage of patients with a URR of 65% or greater compared with the lowest quintile is 0.93/1.19 ⴝ 0.78 (P < 0.0001) for 2002. N ⴝ 2,858 facilities.
ESRD MORTALITY AND FACILITY PRACTICE PATTERNS
131
Fig 5. SMRs for facilities by percentage of patients with a hematocrit of 33% or greater, 1999 and 2002. Categories are defined as approximate quintiles for 1999. The number of facilities in each group is shown at the bottom of each bar. Reference group is facilities with 81.2% to 87.1% of patients with a hematocrit value of 33% or greater in 1999. The relative mortality risk for facilities in the highest quintile of percentage of patients with a hematocrit value of 33% or greater compared with the lowest quintile is 0.94/1.09 ⴝ 0.86 (P < 0.0001) for 2002. N ⴝ 2,858 facilities.
For hematocrit, the fraction of patients achieving K/DOQI guidelines increased by more that 20.6 percentage points at one third of facilities, between 6.0 and 20.6 percentage points for the middle third of facilities, and less than 6.0 percentage points for the final third of facilities. The analogous tertiles for increases in the fraction of patients achieving K/DOQI guidelines with respect to URR were more than 9.4 percentage points, between 0.2 and 9.4 percentage points, and less than 0.2 percentage points. Within the resulting 9 categories of facilities, Table 1 lists average changes in standardized mortality between 1999 and 2002. Improvement in mortality generally is successively greater moving to the right within each row and moving down within each column. This suggests that at each level of improvement in hematocrit (within each row), mortality improvement was greater as URR improvement was greater. Similarly, for each level of URR improvement (within a column), there was more improvement in mortality as improvement in hematocrit was greater. The 359 facilities with the least improvements for URR and hematocrit (upper left cell) experienced a 3% increase in mortality, whereas the 376 facilities
with the most improvements for URR and hematocrit (lower right cell) experienced a 12.4% decline in mortality during this period, on average. Multiple regression analysis shows that compliance with the K/DOQI guidelines for URR and hematocrit have independent effects on mortality. A 10-percentage point increase in fraction of patients with a URR of 65% or greater was associated with a 2.2% decrease in mortality (P ⫽ 0.0006), and a 10-percentage point increase in percentage of patients with a hematocrit value of 33% or greater was associated with a 1.5% decrease in mortality (P ⫽ 0.003) according to regression coefficient estimates. There was no significant interaction between effects of URR and hematocrit slopes on SMR slope (P ⫽ 0.75). Regression diagnostics indicated that the normality assumption was reasonably satisfied, and estimates were not overly influenced by individual data points. DISCUSSION
These results corroborate other studies that have shown increases in hematocrit and dose of dialysis levels for patients with ESRD treated
132
WOLFE ET AL Table 1. Average Percentage of Improvement in Mortality During 1999 to 2002 by Level of Improvement in Practice Patterns Improvement in Percent of Patients With URR ⱖ65%* (no. of facilities)
Improvement in Percent of Patients With Hematocrit ⱖ33%*
None or Decline (⬍0.2%)
Average (0.2%-9.4%)
Large (⬎9.4%)
Small (⬍6.0%) Average (6.0%-20.6%) Large (⬎20.6%)
⫺3% (359) 1.9% (323) 1.2% (262)
1.6% (328) 6.9% (336) 9.7% (306)
2% (257) 12.6% (311) 12.4% (376)
NOTE. Number of facilities (tertiles of hematocrit and tertiles of URR) in each category is shown in parentheses. N ⫽ 2,858 facilities, weighted by average number of patients in the statistic. Total change during 4 years as a percentage of deaths. Positive percentages correspond to lower mortality. *Total percentage point change over 4 years.
with hemodialysis in the United States in recent years. Improvements persisted through the end of 2002. Although there is some variation among types of facility and regions of the country, which are not shown in this report, the improvement largely has been broad based, and the vast majority of facilities have improved K/DOQI compliance for both URR and hematocrit in recent years (Table 1). The national reduction in mortality reported here is encouraging because mortality for patients with ESRD has remained relatively flat during the later years of the 1990s. Although the magnitude of this reduction in mortality is modest, ie, 5.7% between 1999 and 2002, meaning 19 instead of 20 deaths in a typical hemodialysis facility, it represents a reduction of 3,000 deaths/y among more than 60,000 deaths2 annually for US hemodialysis patients. This national reduction in mortality undoubtedly is caused by a wide variety of factors. However, at the facility level, mortality is associated with improvements in both dialysis dose and hematocrit; thus, it is likely that some of this reduction at the national level is associated with improvements in practice patterns. Because this is an observational study, one cannot infer a causal relationship at either the facility level or nationally. Many previous studies documented that mortality is lower among patients receiving higher dialysis doses and those achieving greater hematocrit levels.7-10,21,22 As the norms of standard practice have moved toward higher doses of dialysis, it has become possible to study this relationship at higher doses, and it now is apparent that whereas mortality rates are lower among patients receiving a greater URR, the relation-
ship is flatter at doses greater than a URR of 65% or greater. Port et al21 reported no difference in mortality among males for a URR of 70% or greater, but continued benefits for greater doses among females. These results are consistent with the recent Hemodialysis Study.10 The results here support these prior results and suggest that differences among facilities in achieving K/DOQI guidelines for URR levels are associated significantly with corresponding differences in mortality (Fig 4). Differences in achieved URRs (⬍78% versus ⬎95% at K/DOQI guidelines) and mortality risks (SMR of 1.20 versus 0.92 for 2002) between the worst and best quintiles of facilities are both substantial. These findings do not disagree with the Hemodialysis Study because the Hemodialysis Study did not evaluate patients receiving a URR less than 65%.10 Differences in the fraction of patients achieving a hematocrit value of 33% or greater at the facility level (⬍62% versus ⬎87%) and the corresponding difference in mortality risk (SMR of 1.15 versus 0.94 for 2002) between the worst and best quintiles of facilities also are substantial. These analyses have been adjusted for many patient characteristics; it therefore is unlikely that the relationship shown is caused by unmeasured differences in patient comorbidity. This suggests that K/DOQI goals can be achieved for nearly all patients at most facilities, and the benefits of doing so include longer lifetimes for patients at those facilities. It is possible that changes in URR or hematocrit are, in part, a result of changes in catheter use. During 1999 to 2002, the use of temporary catheters has been decreasing, although there has been increased use of long-term catheters, lead-
ESRD MORTALITY AND FACILITY PRACTICE PATTERNS
ing to an overall increase in catheter use.23 Although dialysis dose and anemia management are unlikely to be independent of other clinical factors, such as catheter use, this study finds an independent association between each of these factors and mortality at the facility level. Facilities that improved dialysis dose, but not hematocrit, for patients still had improvements in mortality, and vice versa. Many facilities improved on only 1 of these 2 factors, not both. Prior facility-based studies have shown that differences in facility practice patterns are associated with differences in mortality.24,25 This is one of the first studies that shows that changes in practice patterns at a facility over time are associated with concurrent changes in mortality. The genesis of the substantial improvement in achieving K/DOQI targets is unclear, but could include the effects of continuous quality improvement efforts by facilities, chains, networks, or the CMS. There have been few studies that document the effect of such changes in anemia and URR practices on such patient outcomes as mortality. As listed in Table 1, facilities that achieved greater improvements in K/DOQI practices also achieved greater reductions in patient mortality. Limitations of this study are as follows. These analyses do not look at patient-by-patient improvement in URR, hematocrit, or mortality, but instead look at aggregates and associations between these aggregates. The associations may not hold at the patient level. We used national Medicare data to study the relationship between practice patterns and mortality at the facility level in 2 ways. First, we studied the relationship of standardized mortality with fractions of patients currently achieving K/DOQI guidelines with regard to URR and hematocrit levels at each facility. Second, we studied the relationship between changes in K/DOQI compliance and changes in mortality using data from 1999 to 2002 to determine whether facilities that improved their K/DOQI compliance levels also had a corresponding decrease in mortality. These analyses were performed at the facility level, which avoids much of the potential bias caused by the healthypatient effect, ie, that patients with greater URR or hematocrit levels may be healthier patients to begin with. The level of achieving practice guidelines was measured at the facility level (as the
133
percentage of patients achieving the guideline), and this measure was used to predict mortality for all patients at that facility. The facility-level measure achieving practice guidelines can be interpreted as a facility-level “intent-to-treat protocol” for all patients at that facility. Facilitylevel analyses do not completely avoid the healthy-patient bias because facilities with a more difficult patient mix are likely to have greater mortality and lower compliance with K/DOQI guidelines because of that patient mix. However, facility-level analyses can reduce this potential bias dramatically. The analysis of changes in this report further reduces this potential for bias because each facility acts as its own control in the calculation of change per year. The facility URR and hematocrit measures studied here were for Medicare patients only, whereas facility mortality was measured for all patients, both Medicare and non-Medicare, at the facilities. This may weaken the associations between mortality and URR and hematocrit, although it is likely that values for Medicare patients at a facility reflect practices for all patients at the facility. One limitation to this study is that patient URR and hematocrit measures are self-reported by facilities. In addition, URR is reported in categories, rather than as a continuous variable. A validation study by Frankenfield et al26 showed that these claims data are a reliable source of information, particularly when classifying patients as above or below threshold values. URR and anemia management may not be independent of other clinical practices in fact or in a statistical sense. Those other practices may (likely do) influence mortality as much as or more than management of these 2 clinical practice parameters. Examples may be the correlation of hemoglobin and albumin levels27 or the use of catheters.28 Unfortunately, information for serum albumin level and vascular access is not available longitudinally in the CMS data. The present study focuses on 2 practices at the dialysis-unit level that may be modified more easily than serum albumin concentration. All observational studies, including this one, are subject to potential bias resulting from unmeasured differences in patient mix that affect both practice measures and patient outcomes. It is plausible that some component of the relation-
134
WOLFE ET AL
ships shown here is caused by unmeasured differences in patient mix among facilities. For example, it might be more difficult to achieve high URR and hematocrit levels at facilities with sicker patients, and those facilities also would be expected to have greater mortality rates. Patient mix at a facility is less likely to change substantially during a 4-year period; therefore, correlations among change scores, as listed in Table 1, are less likely to be affected by such potential bias. Additionally, analyses reported here are adjusted for many patient characteristics, including age and diabetes, which makes it less likely that relationships reported here are caused by unmeasured patient characteristics. In conclusion, the wide variation among facilities in achieving K/DOQI guidelines and that these analyses are adjusted for many patient factors suggests that differences in achievement of K/DOQI guidelines might be caused, in part, by practice pattern differences among facilities, rather than differences in patient mix. The observation that greater improvements in practice patterns, measured by increasing adherence to URR and hematocrit guidelines, correlate with greater improvements in mortality at the facility level strongly suggests benefits from greater adherence to dialysis dose and anemia guidelines. REFERENCES 1. US Renal Data System: USRDS 1999 Annual Data Report. The National Institutes of Health, National Institute of Diabetes and Digestive and Kidney Diseases, Bethesda, MD, 1999 2. US Renal Data System: USRDS 2003 Annual Data Report. The National Institutes of Health, National Institute of Diabetes and Digestive and Kidney Diseases, Bethesda, MD, 2003 3. National Kidney Foundation: K/DOQI Clinical Practice Guidelines for Nutrition in Chronic Renal Failure. Am J Kidney Dis 35:S1-S140, 2000 (suppl 2) 4. Centers for Medicare and Medicaid Services: 2002 Annual Report, End-Stage Renal Disease Clinical Performance Measures Project. Baltimore, MD, Department of Health and Human Services, Centers for Medicare and Medicaid Services, Center for Beneficiary Choices, 2002 5. Wolfe RA, Ashby VB, Milford EL, et al: Comparison of mortality in all patients on dialysis, patients on dialysis awaiting transplantation, and recipients of a first cadaveric transplant. N Engl J Med 341:1725-1730, 1999 6. US Organ Procurement and Transplantation Network and the Scientific Registry for Transplant Recipients: 2002 Annual Report of the US Organ Procurement and Transplantation Network and the Scientific Registry for Transplant
Recipients: Transplant Data 1992-2001: Department of Health and Human Services, Health Resources and Services Administration, Office of Special Programs, Division of Transplantation, Rockville, MD; United Network for Organ Sharing, Richmond, VA; University Renal Research and Education Association, Ann Arbor, MI, 2002 7. Wolfe RA, Ashby VB, Daugirdas JT, Agodoa LY, Jones CA, Port FK: Body size, dose of hemodialysis, and mortality. Am J Kidney Dis 35:80-88, 2000 8. Port FK, Ashby VB, Dhingra RK, Roys E, Wolfe RA: Dialysis dose and body mass index are strongly associated with survival in hemodialysis patients. J Am Soc Nephrol 13:1061-1066, 2002 9. Collins AJ, Li S, St Peter W, et al: Death, hospitalization, and economic associations among incident hemodialysis patients with hematocrit values of 36 to 39%. J Am Soc Nephrol 12:2465-2473, 2001 10. Eknoyan G, Beck GJ, Cheung AK, et al, the Hemodialysis (HEMO) Study Group: Effect of dialysis dose and membrane flux in maintenance hemodialysis. N Engl J Med 347:2010-2019, 2002 11. Paniagua R, Amato D, Vonesh E, et al: Effects of increased peritoneal clearances on mortality rates in peritoneal dialysis: ADEMEX, a prospective, randomized, controlled trial. J Am Soc Nephrol 13:1307-1320, 2002 12. Wolfe RA, Gaylin DS, Port FK, Held PJ, Wood CL: Using USRDS generated mortality tables to compare local ESRD mortality rates to national rates. Kidney Int 42:991996, 1992 13. SAS Institute Inc: SAS/STAT User’s Guide, version 8. Cary, NC, SAS Institute Inc, 1999, pp 2598-2599 14. Andersen PK, Borgun O, Gill RD, Keiding N: Statistical Models Based on Counting Processes. New York, NY, Springer-Verlag, 1993, pp 334, 406-407 15. Collett D: Modeling Survival Data in Medical Research. London, UK, Chapman & Hall, 1994, p 153, equation 5.6, p 151, equation 5.1 16. Wolfe RA, Ashby VB, Port FK: 1993 DMMS comorbidity index validated by Medical Evidence Form data. J Am Soc Nephrol 11:1300A, 2000 (abstr) 17. National Center for Health Statistics. Health, United States, 2002 With Rural and Urban Health Chartbook. Hyattsville, MD, Centers for Disease Control and Prevention, Health and Human Services Department, 2002, pp 117-118 18. Wolfe RA, Ashby VB, Hulbert-Shearon TE, Port FK: New dialysis facility mortality statistics (SMRs) adjust for more patient characteristics. J Am Soc Nephrol 12:1802A, 2001 (abstr) 19. Wolfe RA, Ashby VB, Port FK, et al: The association between regional death rates and standardized mortality ratios (SMR) by health service areas (HSA) and race and sex group. J Am Soc Nephrol 9:230A, 1998 (abstr) 20. Ashby VB, Wolfe RA, Loos ME, Port FK: The effect of comorbidities on facility standardized mortality ratios. J Am Soc Nephrol 9:197A, 1998 (abstr) 21. Port FK, Wolfe RA, Hulbert-Shearon TE, McCullough KP, Ashby VB, Held PJ: High dialysis dose is associated with lower mortality among females but not among males. Am J Kidney Dis 43:1014-1023, 2004
ESRD MORTALITY AND FACILITY PRACTICE PATTERNS
22. Ma J, Ebben J, Xia H, Collins A: Hematocrit level and associated mortality in hemodialysis patients. J Am Soc Nephrol 10:610-619, 1999 23. Centers for Medicare and Medicaid Services: 2002 Annual Report, End-Stage Renal Disease Clinical Performance Measures Project. Baltimore, MD, Department of Health and Human Services, Centers for Medicare and Medicaid Services, Center for Beneficiary Choices, 2003 24. Lowrie EG, Teng M, Lacson E, Lew N, Lazarus JM, Owen WF: Association between prevalent care process measures and facility specific mortality rates. Kidney Int 60:1917-1929, 2001 25. McClellan WM, Soucie JM, Flanders WD: Mortality in end-stage renal disease is associated with facility-to-
135
facility differences in adequacy of hemodialysis. J Am Soc Nephrol 9:1940-1947, 1998 26. Frankenfield DL, Brier ME, Bedinger MR, et al: Comparison of urea reduction ratio and hematocrit data reported in different data systems: Results from the Centers for Medicare and Medicaid Services and the Renal Network Inc. Am J Kidney Dis 41:433-441, 2003 27. Pisoni RL, Bragg-Gresham JL, Young EW, et al: Anemia management outcomes from 12 countries in the Dialysis Outcomes and Practice Patterns Study (DOPPS). Am J Kidney Dis 44:94-111, 2004 28. Dhingra RK, Young EW, Hulbert-Shearon TE, Leavey SF, Port FK: Type of vascular access and mortality in US hemodialysis patients. Kidney Int 60:1443-1451, 2001