EVALUATING NKF-DOQI GUIDELINES 785 mortality ratio to compare mortality rates among dialysis units, adjusting for age, sex, race, and diabetic status.

Transcription

1 United States Renal Data System Assessment of the Impact of the National Kidney Foundation-Dialysis Outcomes Quality Initiative Guidelines Allan J. Collins, MD, Tricia L. Roberts, MS, Wendy L. St. Peter, PharmD, Shu-Cheng Chen, MS, Jim Ebben, BS, and Edward Constantini, MA Since 1989, significant efforts have focused on improving the care of dialysis patients in the United States. Numerous organizations have developed clinical practice guidelines; however, few guidelines have received the broad support given to the National Kidney Foundation-Dialysis Outcomes Quality Initiative (DOQI). These guidelines, independently developed from an extensive review of the literature, include sections on dialysis adequacy, anemia treatment, and vascular access. To assess the impact of these guidelines on clinical practice, we evaluated data on hematocrits, recombinant human erythropoietin dosing, hemodialysis adequacy, and simple fistula and dialysis catheter utilization using Medicare dialysis provider claims and Medicare Part B physician services. Hematocrits have increased steadily, with the exception of the period when the Hematocrit Measurement Audit was in effect. After cancellation of the policy, hematocrits increased to the midpoint of the DOQI target range (34.4%). Although the level of dialysis therapy has stabilized, with the average urea reduction rate of 68% to 69.9% in 1997 to 1999 being slightly greater than the DOQI target of 65% or greater, geographic variability is apparent. Simple fistula placement rates increased by 45% during the pre-doqi and post-doqi period from 1994 to The use of temporary catheters decreased, whereas placement of permanent catheters has increased, which may reflect recommended practice guidelines. Although it appears that clinical practice guidelines have improved the clinical care of dialysis patients, considerable regional variations in care across the country should be given significant attention by the National Kidney Foundation, Inc. INDEX WORDS: End-stage renal disease (ESRD); National Kidney Foundation-Dialysis Outcomes Quality Initiative (NKF-DOQI) guidelines; hemodialysis (HD); hematocrit; catheters; arteriovenous fistulae; dialysis adequacy; clinical practice guidelines. From the US Renal Data System and US Renal Data System Information System, Minneapolis, MN. Received March 22, 2001; accepted in revised form October 26, Funded in whole with federal funds by contract no. N01-DK from the National Institute of Diabetes and Digestive and Kidney Diseases, The National Institutes of Health. Address reprint requests to Allan J. Collins, MD, US Renal Data System, Minneapolis Medical Research Foundation, 914 S 8th St, Ste D-206, Minneapolis, MN acollins@usrds.org 2002 by the National Kidney Foundation, Inc /02/ $35.00/0 doi: /ajkd Editorial, p. 889 IN 1989, THE DALLAS Symposium on Morbidity and Mortality of dialysis patients identified major differences in patient survival among the United States, Europe, Canada, Australia, and Japan. 1 The United States had the highest rates of newly treated patients with end-stage renal disease (ESRD), but also, at 22% to 24%, the highest gross mortality rate. Conversely, Japan s mortality rate was only 8% to 9%. Conference participants focused on dialysis therapy as a major area of concern in the United States, noting that shorter treatment times, reuse of dialyzers, and use of dialysis catheters and grafts for vascular access may be sources of differences. In the 7 years after the conference, efforts to improve the quality of dialysis care in the United States were undertaken by the Health Care Financing Administration (HCFA; now the Centers for Medicare and Medicaid Services [CMS]), 2 the Association for the Advancement of Medical Instrumentation (AAMI), 3 the Renal Physicians Association (RPA), and The National Institutes of Health (NIH). These efforts culminated in the Dialysis Outcomes Quality Initiatives (DOQI) project, organized by the National Kidney Foundation (NKF). 4 Each of these organizations contributed consensus documents based on their own work that were directed at improving some element of the care of dialysis patients. HCFA, for example, began the ESRD Core Indicators Project, tracking hematocrits, dialysis therapy, iron saturation, and ferritin levels, and blood pressure control. 2 In 1992, Wolfe et al 5 developed the standardized 784 American Journal of Kidney Diseases, Vol 39, No 4 (April), 2002: pp

2 EVALUATING NKF-DOQI GUIDELINES 785 mortality ratio to compare mortality rates among dialysis units, adjusting for age, sex, race, and diabetic status. The AAMI used surveys of dialysis units to revise and improve standards for reuse practices. 3 The NIH convened a conference on dialysis care that recommended improvements in, as well as assessment of, the quality of life of dialysis patients, and the RPA promoted a Kt/V of 1.2 for effective dialysis therapy. 6 In 1995, the NKF, through its NKF-DOQI initiative, created work groups to review the medical literature in a scientifically rigorous manner and develop detailed practice guidelines in the areas of anemia, dialysis adequacy, and vascular access. The full set of guidelines, published in fall 1997, 4 included, for dialysis patients, a target hematocrit of 33% to 36%, along with a minimum Kt/V of 1.2 (single pool) and/or a urea reduction ratio (URR) of 65% or greater per dialysis treatment. The group recommended that to reduce morbidity related to vascular access, the use of simple fistulas should be increased from current levels of 10% to 15% (compared with 50% in other parts of the world) and suggested that complications may be reduced through the use of internal jugular vein cuffed catheters, rather than subclavian temporary catheters. Since the introduction of the NKF-DOQI guidelines in 1997, there has been little largescale assessment of their impact on clinical care. The HCFA ESRD Clinical Performance Measures Project (formerly the ESRD Core Indicators Project) routinely assesses dialysis adequacy, hematocrits, and the percentage of patients administered iron therapy on a network level. Because this project assesses data on only 6,000 to 8,000 patients a year, it provides little information on geographic variations in clinical practices. Our study is designed to use Medicare data to independently assess the impact of the NKF- DOQI guidelines on anemia management, hemodialysis adequacy, use of dialysis catheters, and placement rates of simple fistulas. METHODS To assess the NKF-DOQI guidelines, we constructed data sets from the available coded information in the HCFA billing system. Each method is directed at a specific guideline. Anemia Management: NKF-DOQI Guideline 4, Target Hemoglobin/Hematocrit NKF-DOQI guideline 4 of the anemia guidelines suggests a target hemoglobin range for recombinant human erythropoietin (rhuepo) therapy of 11 to 12 g/dl (hematocrit, 33% to 36%). Data Source Hematocrit data were obtained from institutional outpatient Medicare claims for rhuepo-treated patients with ESRD from January 1, 1994, to April 30, Data on rhuepo dose and hematocrit are included on each claim, along with the number of doses administered (service units). Claims with hematocrit values less than 10% or greater than 50%, as well as those reporting greater than 20 service units of rhuepo in a month, were excluded from all analyses because of probable errors in data. For each patient, the monthly mean hematocrit was determined from all hematocrit claims within the month. Mean rhuepo dose per administration was calculated by summing all doses and dividing by the total number of service units within the billing period. Mapping of Data Maps showing geographic patterns in hematocrits and rhuepo doses were constructed on the level of health service area (HSA) 7 for 1999 using Medicare claims data for period prevalent hemodialysis patients administered rhuepo. Each patient s mean hematocrit and mean rhuepo dose per administration for the year were calculated from claims dated from the latest of January 1 or day 91 of ESRD until December 31, To assess how many patients achieved the lower limit of the hematocrit range recommended by the DOQI, the percentage of patients with hematocrits of 33% or greater was determined for each HSA. These data then were smoothed using the weighted head-banging method. 8,9 rhuepo dose and hematocrit data were weighted by the total number of 1999 rhuepo administrations and total number of patients with rhuepo claims per HSA, respectively. Smoothed HSA-level values then were divided into quintiles and mapped with color variations to illustrate geographic patterns. Three-Month Rolling Average Hematocrit Condensed Medicare claims from rhuepo-treated patients with ESRD were used to obtain hematocrit measurements from November 1, 1995, to April 30, Only patients with hematocrit values (rhuepo claims) for 3 consecutive months (the current and 2 previous months) were included in each monthly analysis, and a 3-month rolling average hematocrit was calculated for each patient from January 1996 to April For each month, percentages of patients with a 3-month rolling average hematocrit within the following ranges were determined and graphed as a sand diagram, a layered distribution graph of hematocrit groups across time: less than 30%, 30% to less than 33%,

3 786 33% to less than 36%, 36% to less than 37.5%, 37.5% to less than 39%, and 39% or greater. Hematocrit Variability In further analyses, we explored the variability in monthly hematocrit data by determining the percentage of patients with a 3-month rolling average hematocrit within the DOQI target range and determining the number of monthly hematocrit values exceeding the target range of 33% to less than 36% from January 1996 to April Analyzing the three monthly hematocrit values contributing to each 3-month rolling average, we calculated the percentage of patients with no, one, or two monthly hematocrit values of 36% or greater and 39% or greater for each month, with sand diagrams constructed to show the distribution of patients across time. Hemodialysis Adequacy: NKF-DOQI Guideline 4, Minimum Delivered Dose of Hemodialysis Guideline 4 of the NKF-DOQI hemodialysis adequacy guidelines recommends a minimum delivered dose Kt/V of 1.2, corresponding to an average URR of 65%. URR Average URR data from 1994 to 1999 and percentage of change in average URRs from 1994 for adult in-center hemodialysis patients were obtained from the 1998 HCFA ESRD Core Indicators Project and 1999 to 2000 HCFA Clinical Performance Measures Project reports. 2,10,11 URR data for 1999 were obtained from Medicare institutional outpatient dialysis unit claims with revenue codes 821 or 825, a Physicians Current Procedural Terminology (CPT) code of 90999, and a G modifier with a value of G1 to G5. Modifier values of G1 to G5 indicated the following URR ranges: G1, less than 60%; G2, 60% to 64.9%; G3, 65% to 69.9%; G4, 70% to 74.9%; and G5, 75% or greater. Patients for whom fewer than seven dialysis sessions were provided in a month were excluded because a G6 modifier and no URR range were provided for these patients. It is not possible to determine with certainty which patients receive regular twice- or thrice-weekly dialysis therapy within Medicare claims data because dialysis claims are not generated when a patient is hospitalized. The median URR range in 1999 was determined for each patient; for patients with an even number of URR values, the two middle values were selected, and each was weighted by one half. Within each HSA, the percentage of patients with a median URR within each URR range was calculated. Using the weighted headbanging method, data were smoothed and weighted by the number of 1999 prevalent hemodialysis patients in each HSA. A map was created of the percentage of patients with median 1999 URRs of 65% or greater. Vascular Access: NKF-DOQI Guideline 3: Selection of Permanent Vascular Access This guideline recommends primary arteriovenous (AV) fistulas as the preferred type of vascular access, followed by AV grafts, and discourages the use of cuffed tunneled central venous catheters for permanent access. COLLINS ET AL Fistula Procedure Rates Primary Medicare period prevalent hemodialysis patients for 1994 to 1999 were studied, and fistulas were identified from Medicare Part B physician/supplier claims (CPT codes: 36819, transposition vein; 36821, simple fistula; and 36825, autogenous vein graft). Duplicate occurrences of the same CPT code with the same first and last expense date on multiple lines of a claim were counted as one fistula creation. To determine the exposure period on hemodialysis therapy, start of follow-up was defined as January 1 of the year for prevalent patients and day 91 of ESRD for incident patients, whereas the end of follow-up was the earliest of death, modality change, loss to follow-up, or December 31 of the year. To assess the placement of AV fistulas, complete service claims from the Medicare system were required. Patients failing to reach a certain level of Medicare paid dialysis claims were excluded from fistula rate calculations because these patients were likely to have Medicare as a secondary payer (MSP) and would contribute years at risk with incomplete event data. Patients included in the analysis had a start of follow-up between the following dates: (1) the first day with Medicare paid dialysis claims in a month in which there was at least $675 in dialysis claims, and (2) the end of a 3-month period in which there was less than $675 of Medicare paid dialysis claims in each month. If the date of the start of follow-up did not fall between these dates, the patient was excluded from the analysis for that year because billing data were incomplete. 12 Patients with official MSP classification, described by HCFA in its 1993 to 1995 ESRD research report, also were excluded from the analysis. 13 For each year, a patient s time at risk was calculated as the time from the start to the end of follow-up, and total patient-years at risk were summed. For each year, gross fistula rates per 1,000 patient-years at risk were calculated as the total number of fistula codes divided by total patientyears at risk multiplied by 1,000. The fistula creation rate per 1,000 patient-years at risk was calculated for each HSA for These data were smoothed by the weighted head-banging method and weighted by the total number of period prevalent hemodialysis patients in the HSA for the year. A map then was created showing geographic patterns in fistula creation rates. Vascular Access: NKF-DOQI Guidelines 5 and 6: Tunneled Cuffed and Noncuffed Catheters Guideline 5 of the NKF-DOQI vascular access guidelines promotes tunneled cuffed venous catheters for temporary access of longer than 3 weeks duration, whereas guideline 6 suggests the use of noncuffed catheters for access of less than 3 weeks duration. Catheter Placement Rates Using 1994 to 1999 period prevalent hemodialysis patients, unadjusted temporary and permanent central venous catheter insertion rates per 1,000 patient-years at risk were calculated. Medicare Part B physician/supplier claims with CPT codes of 36489, 36491, and (indicating tempo-

4 EVALUATING NKF-DOQI GUIDELINES 787 rary catheters) and a CPT code of (indicating permanent catheters) were used to determine placement of a catheter. Rate calculation and mapping methods follow those used in the fistula rate calculation, including the removal of MSP patients based on failure to reach a certain level of Medicare paid dialysis claims or official MSP status as noted in Part A and Part B claims data. Because central catheters may be placed for indications other than dialysis, additional methods were used to identify catheters used for dialysis. Part B physician/supplier and durable medical equipment claims for 1994 to 1999 were searched for chemotherapy (CPT codes 96408, 96410, and 96412) and parenteral nutrition (CPT codes B4164-B5200, B9004, B9006, and B9999) claims. Patients with at least one of these codes during the year were excluded for the year. Also, because CPT codes 36533, 36489, and are not specific to dialysis, International Classification of Diseases, Ninth Revision (ICD-9) diagnosis codes at both the line and claim levels were analyzed to exclude catheter insertions having the previously noted CPT codes, along with non renal failure diagnosis codes. A CPT code of 36533, 36489, or was included if it was associated with either a line-level diagnosis code or a claim-level principal diagnosis code among the following ICD-9 codes related to dialysis or renal failure: 250, 403, 580 to 589, 593, 996.1, , , V45.1, and V56. Permanent Catheter-Day Rates Unadjusted rates of permanent catheter-days per patientyear at risk and days per catheter insertion were calculated for 1994 to 1999 period prevalent hemodialysis patients. This method followed that used to calculate catheter rates, incorporating the exclusion of MSP patients, exclusion of patients with chemotherapy or parenteral nutrition, and exclusion of catheter insertions (CPT code 36533) without a renal failure, dialysis claim-level principal diagnosis code, or line-level diagnosis code. Also following the method for fistula and catheter rates, time at risk was calculated from the latest of day 91 of ESRD or January 1 of the year until the earliest of death, modality change, loss to follow-up, or December 31 of the year. Permanent catheter-days were counted during the time at risk by summing days from a catheter insertion (CPT code 36533) until a catheter removal (CPT code 36535). Because catheter insertions and removals occur on claims on separate lines with different expense dates, it was possible for two insertions or two removals to occur consecutively. In this case, days were counted from the first insertion or until the last removal, respectively. When the first code during the time at risk indicated a removal, permanent catheter-days were counted from the beginning of the time at risk until the removal. Similarly, in the case of an insertion as the last code of the time at risk, catheter-days were counted from the insertion until the end of the risk period. An insertion and removal occurring on the same day were counted as one catheter-day. Rates of permanent catheter-days per year at risk then were calculated by dividing total catheter-days during the time at risk by total years at risk, and the number of days per insertion was computed by dividing total days by the number of insertions during the time at risk. RESULTS Anemia Management: NKF-DOQI Guideline 4, Target Hemoglobin/Hematocrit rhuepo and Hematocrit Trends Figure 1 shows the mean hematocrit and mean rhuepo dose per administration by month from January 1994 to April In general, hematocrits increased progressively until the Hematocrit Measurement Audit (HMA) was put into effect by HCFA, at which point mean hematocrits began to decrease. Modification and elimination of the HMA policy in 1998 was accompa- Fig 1. Mean US hematocrit (Hct) and mean rhuepo (EPO) dose per administration each month in prevalent dialysis patients with rhuepo claims, January 1994 to April 2000.

5 788 COLLINS ET AL Fig 2. (Top) Geographic variability in mean hematocrit (%) and (bottom) percentage of patients meeting the lower limit of the DOQI target hematocrit range (>33%) in prevalent hemodialysis patients with rhuepo claims in 1999 by HSA. Each color represents quintiles of HSAs with mean hematocrits and percent of patients meeting the DOQI target. nied by a steady increase in mean hematocrits until the end of 1999 and beginning of 2000, when hematocrits appeared to stabilize between 34% and 34.5%. At the same time, monthly mean rhuepo doses appeared to stabilize at just greater than 5,900 U/administration. In April 2000, mean hematocrit was 34.4% and mean rhuepo dose was 5,923 U/administration. The greatest mean hematocrits ( 34.6%) and lowest mean rhuepo doses ( 5,150 U/administration) occurred in the Pacific Coast, Rocky Mountain, and some Southwestern and Northeastern states, as well as Minnesota (Figs 2 and 3). The greatest percentages of patients meeting the lower limit of the DOQI guidelines (hematocrit 33%) were located in the same regions (Fig 2). Generally, states in the Central, Southeast, and Eastern portions of the United States Fig 3. Average rhuepo dose per administration in prevalent hemodialysis patients with rhuepo claims in 1999 by HSA. Each color of the map represents quintiles of HSAs and the mean rhuepo dose within the quintiles.

6 EVALUATING NKF-DOQI GUIDELINES 789 showed the lowest mean hematocrits ( 33.8%) and highest mean rhuepo doses ( 6,180 U/administration). For mean hematocrits shown in Fig 2, the hematocrit intervals chosen ( 33.8%, 33.8% to 34.0%, 34.0% to 34.3%, 34.3% to 34.6%, and 34.6 %) appear narrow because they are based on quintiles of the HSA-level mean hematocrit values, rather than individual patient-level hematocrit measurements. With sufficient sample size, even a small difference in mean hematocrit may be very meaningful and clinically relevant, especially when distinct geographic trends are evident. Three-Month Rolling Average Hematocrit Three-month rolling average hematocrit values increased consistently after the release of the DOQI guidelines and discontinuation of the HMA prepayment review (Fig 4). The percentage of patients with 3-month rolling average hematocrits of 36% or greater increased from 4.1% in January 1996 to 10.2% in June 1998 and to 29.9% in April During the last quarter of 1999 and first quarter of 2000, the percentage of patients with 3-month rolling average hematocrits less than 33% and 36% or greater stabilized, with approximately 30% of patients falling within each of these groups. Hematocrit Variability The percentage of patients with a rolling 3-month average hematocrit of 33% to less than 36% and with one or more monthly hematocrit values of 36% or greater or 39% or greater increased over time from 33.8% in January 1996 to 62.6% in April 2000 and from 3.2% in January 1996 to 8.9% in April 2000, respectively (Fig 5). The growing percentage of patients with monthly hematocrits of 36% or greater or 39% or greater followed the elimination of the HMA in July This trend also appeared to stabilize in the last quarter of 1999 and first quarter of Hemodialysis Adequacy: NKF-DOQI Guideline 4, Minimum Delivered Dose of Hemodialysis URR The average URR for adult in-center hemodialysis patients increased from 63.8% in 1994 to 69.9% in 1999 (Fig 6). The greatest percentage of patients meeting the DOQI target URR of 65% or greater lived in the Rocky Mountain region, Texas, and Alaska (Fig 7). Vascular Access: NKF-DOQI Guideline 3: Selection of Permanent Vascular Access Fistula Procedure Rates For each year s data, 21% to 27% of the period prevalent hemodialysis patients were omitted from the analysis based on MSP classification. Numbers of remaining patients with Medicare as the primary payer are listed in Table 1. Figure 8 shows trends in fistula rates and temporary catheter and permanent catheter rates for 1994 to An overall rate of 71.1 fistula creations/1,000 patient-years at risk occurred in 1999, a 13.4% increase from the 1998 rate. The greatest fistula rates occurred in areas of the Northeastern, Central, and Western United States (Fig 9). Fig 4. Percentage of patients by hematocrit group (see key) using a calculated 3-month rolling average hematocrit, January 1996 to April 2000.

7 790 COLLINS ET AL Fig 5. Patients with a 3-month rolling average hematocrit of 33% to less than 36%: Percentage of patients with 0, 1, or 2 months with hematocrits of (top) 36% or greater and (bottom) 39% or greater, January 1996 to April Vascular Access: NKF-DOQI Guidelines 5 and 6: Tunneled Cuffed and Noncuffed Catheters Catheter Placement Rates One percent of non-msp patients were excluded because of at least one chemotherapy or parenteral nutrition claim during the year. Total patients included each year in the catheter rate and catheter-day rate calculations are listed in Table 1. Approximately 20% of the total 1994 to 1999 temporary and permanent catheter inser- Fig 6. Average URR of US adult in-center hemodialysis patients. Patients are included regardless of insurance status. (Data from the 1998 HCFA ESRD Core Indicators Project and 1999 to 2000 HCFA Clinical Performance Measures Project reports, shown for October through December of each year.)

8 EVALUATING NKF-DOQI GUIDELINES 791 Fig 7. Percentage of prevalent hemodialysis patients with a URR of 65% or greater in 1999 by HSA. Each color represents quintiles of HSAs with the percent of patients meeting the URR target. tions were omitted from analyses because the primary diagnosis code on the CPT code of 36533, 36489, or was for a nonrenal principal diagnosis. From 1994 to 1999, permanent central venous catheter insertion rates increased 181% (from 74 to 207 insertions/1,000 patient-years at risk) and temporary central venous catheter rates decreased 19% (from 490 to 397; Fig 8). In 1999, temporary catheter insertion rates were generally highest ( 421) in the Southeast (Fig 9). Areas of the Southeastern and Western United States also showed the greatest ( 229) permanent catheter placement rates; however, these areas did not exactly correspond with areas with the greatest temporary catheter insertion rates. Permanent Catheter-Day Rates Figure 10 shows 1994 to 1999 rates for permanent catheter-days, with a rate of 148 d/insertion in This figure also shows an overall 1999 rate of 30.7 d/patient-year at risk, with rates increasing 167% from 1994 to Areas of lowest permanent catheter-days per insertion ( 120 days) appear in the Midwest in 1999 (Fig 11). DISCUSSION As shown through our data, the development of clinical practice recommendations and guidelines by the major organizations responsible for dialysis patient care has had a clear impact on provider practices. For instance, both the HCFA Clinical Performance Measures Project and trend analysis have shown increases in hematocrits and dialysis therapy. 2,14 The target URR was proposed by several organizations before the NKF-DOQI guidelines were published, and it appears that providers were responding to these consensus recommendations and guidelines before The recent addition of URR ranges on dialysis provider claims has provided an opportunity to determine not only the reported amount of dialysis therapy, but also the degree of geographic variation (Fig 7). The fewest patients with a URR of 65% or greater (the lower limit of the DOQI target) appeared in areas in California, Oregon, Nevada, Utah, Montana, North Dakota, Iowa, Nebraska, Kansas, Illinois, Oklahoma, Louisiana, Kentucky, West Virginia, and other areas of the Atlantic coast states in Table 1. Number of Period Prevalent Hemodialysis Patients Included in Fistula Rate, Catheter Rate, and Catheter-Days Analyses, 1994 to 1999 Year No. of Patients No. After Exclusion of MSP Patients* No. After Exclusion of Patients With Chemotherapy or Parenteral Nutrition Claims , , , , , , , , , , , , , , , , , ,753 *Equals size of study population included in fistula rate analyses. Equals size of study population included in catheter rate and days analyses.

9 792 COLLINS ET AL Fig 8. Insertion rates for central venous catheters and creation rates for simple fistulas per 1,000 patientyears at risk in US prevalent hemodialysis patients, 1994 to The NKF-DOQI target URR is based on the assumption that hemodialysis sessions are delivered three times weekly. Unfortunately, it is not possible to delineate between patients who undergo hemodialysis two or three times weekly using Medicare claims data. Patients who undergo dialysis while they are hospitalized will not generate a dialysis claim. Thus, it is possible that patients who receive twice-weekly dialysis treatments may have been included in the geographic URR analysis, and some patients who received thrice-weekly dialysis treatment may have been excluded. However, this analysis was primarily meant to point out that there is tremendous geographic variation in URRs across the country. Many areas with low dialysis therapy also have greater dialysis catheter placement rates (Fig 9), suggesting that persistent low levels of therapy are clinically and epidemiologically associated with the use of dialysis catheters. Because dialysis therapy already was increasing before the NKF-DOQI guidelines, this initiative appears to have had a minimal impact. The plateau in URRs in 1997 and 1998 should be reevaluated now that reporting URRs is required on dialysis unit claims. However, that providers may report the highest URR in a month should be taken into consideration. In addition, because accuracy in predialysis and postdialysis sampling of blood can influence results, sampling techniques and URR inflation need to be carefully assessed. Despite these concerns, providers across the country appear to be delivering more dialysis therapy. Anemia treatment has been influenced more clearly by government payment policies and the DOQI target range. Although hematocrits increased steadily after 1990, 14 the implementation of HCFA s HMA policy in fall 1997 had a dramatic impact on hematocrit levels. Under this policy change, fiscal intermediaries (FIs) and carriers were to process rhuepo claims based on a 90-day (3-month) average hematocrit when the current month s hematocrit was greater than 36%. This event triggered the HMA, at which point the FI or carrier was to average the current month s hematocrit with the 2 preceding months. The rhuepo claim was to be denied if this average rolling 3-month hematocrit was greater than 36.5%. This policy achieved the goal of reducing the percentage of patients with hematocrit levels greater than 36%. Unfortunately, providers fear of the loss of rhuepo reimbursement led to decreasing or holding rhuepo doses. Ultimately, the most concerning impact was on patients because decreasing rhuepo doses led to decreasing mean hematocrits (less than the suggested target hematocrit range) in almost all hematocrit groups. Consequently, the HMA policy was shown to be in conflict with both the NKF-DOQI Clinical Guidelines and HCFA s ESRD Core Indicator Project (now renamed the Clinical Performance Measure Project), and the policy was modified and then rescinded in Current HCFA policy instructs FIs to conduct a postpayment review of rhuepo claims and identify providers with atypical numbers of patients with 90-day average hematocrits greater than 37.5%. Currently, the definition of atypical has been left to the discretion of the FIs. The percentage of patients with 3-month rolling average hematocrits of 37.5% or greater increased after the HMA was rescinded. However, our data show that this upward trend appears to have stabilized, with 14.6% of patients having a 3-month rolling average hematocrit of 37.5% or greater in April Some of these patients

10 Fig 9. (Top) Fistula creation rates and (middle) permanent and (bottom) temporary central venous catheter insertion rates per 1,000 patient-years at risk in prevalent hemodialysis patients in 1999 by HSA. Each color represents quintiles of HSAs with the insertion rates. Fig 11. Permanent catheter-days per insertion in prevalent hemodialysis patients in 1999 by HSA. Each color represents quintiles of HSAs by the mean catheter days.

11 794 COLLINS ET AL Fig 10. Permanent catheter-days per year at risk and per insertion in US prevalent hemodialysis patients, 1994 to most likely had medical justification for higher hematocrits. It is interesting to note that even when a 3-month rolling average hematocrit of 33% to 36% is achieved, 60% of patients have at least one hematocrit in the 3-month period greater than the target range (Fig 5). Ultimately, the average hematocrit stabilized in the midpoint of the DOQI target range, suggesting that providers have responded to the DOQI guidelines. However, there is considerable geographic variation in percentages of patients meeting the DOQI target range, and several areas with the highest mean hematocrits have the lowest mean rhuepo doses. Iron therapy was not assessed at this time, but should be given attention in the future because variations in iron use across the United States may help explain this interesting observation. In addition, some areas of the country may have a greater percentage of hemodialysis patients administered subcutaneous rhuepo, which results, on average, in lower rhuepo doses than when administered intravenously. The Medicare database does not allow distinction between intravenous and subcutaneous dosing of rhuepo. Although the use of subcutaneous administration in this group of patients is not very high compared with peritoneal dialysis patients, it also may explain why some regions have lower mean rhuepo doses, but higher mean hematocrits. The impact of the DOQI guidelines on vascular access practices also is clear. From 1994 to 1999, placement rates of simple fistulas increased, with the steepest increase occurring after Compared with 1996, placement rates increased 35% in The use of temporary dialysis catheters has decreased annually from 1994 to However, temporary catheter insertion rates decreased only slightly from 1994 to 1996, showing a greater decline after the release of the DOQI guidelines. Temporary catheter insertion rates decreased 17% from 1996 to Conversely, permanent catheter insertion rates increased fairly consistently both before and after the DOQI guidelines, with a 71% increase from 1996 to The guideline to reduce the use of temporary catheters appears to have come after providers began to expand the use of permanent internal jugular catheters. The transition from temporary to permanent catheters also is reflected in geographic data. The greatest rates of permanent catheter placements occurred in the Midnorthern and Southern part of the United States and the Ohio River Valley area, as well as in the mid-atlantic states. The number of days per permanent catheter insertion has not changed much since 1994, but the total number of catheter-days per patient-year at risk has increased 167% from 1994 to 1999 (Fig 10). In summary, peer-reviewed publications, which provided the basis for the development of recommendations and guidelines for dialysis patient care by various governmental and organizational bodies, along with monitoring care, have led to changes in clinical practices. The lower incident death rates reported in the US Renal Data System 2000 Annual Data Report occurred as the quality of care was increasing. 12 A more direct assessment of trends in mortality and hospitalization would establish the magnitude of these associations and determine whether guidelines continue to have an impact on clinical practice. ACKNOWLEDGMENT The authors thank HCFA (now the Centers for Medicare and Medicaid Services [CMS]) for their assistance with CMS data; Beth Forrest and Dana Knopic for help in

12 EVALUATING NKF-DOQI GUIDELINES 795 manuscript preparation and regulatory assistance; and C. Daniel Sheets for providing data set construction and data manipulation. REFERENCES 1. Hull A, Parker T: Proceedings from the Morbidity, Mortality, and Prescription of Dialysis Symposium, Dallas, Texas, September 15 to 17, Am J Kidney Dis 15: , Health Care Financing Administration: 1999 Annual Report, End-Stage Renal Disease Clinical Performance Measures Project. Baltimore, MD, Department of Health and Human Services, HCFA, Office of Clinical Standards and Quality, Association for the Advancement of Medical Instrumentation: AAMI Standards and Recommended Practices, volume 3, Dialysis. Arlington, VA, AAMI, Eknoyan G, Levin N: Clinical Practice Guidelines: Final Guideline Summaries from the Work Groups of the National Kidney Foundation-Dialysis Outcomes Quality Initiative. New York, NY, National Kidney Foundation, Wolfe R, Gaylin D, Port F, Held P, Wood CL: Using USRDS generated mortality tables to compare local ESRD mortality rates to national rates. Kidney Int 42: , Renal Physician s Association: Clinical Practice Guideline on Adequacy of Hemodialysis. Washington, DC, Renal Physician s Association, Pickle LW, Mungiole M, Jones GK, White H: Atlas of United States Mortality. Hyattsville, MD, National Center for Health Statistics, Clayton D, Kaldor J: Empirical Bayes estimates of age-standardized relative risks for use in disease mapping. Biometric 43: , Manton KG, Woodbury M, Stallard E, Riggan WB, Creason JP, Pellom AC: Empirical Bayes procedures for stabilizing maps of US cancer mortality rates. J Am Stat Assoc 84: , Health Care Financing Administration: 1998 Annual Report, End-Stage Renal Disease Core Indicators Project. Baltimore, MD, Department of Health and Human Services, HCFA, Office of Clinical Standards and Quality, Health Care Financing Administration: 2000 Annual Report, End-Stage Renal Disease Clinical Performance Measures Project. Baltimore, MD, Department of Health and Human Services, HCFA, Office of Clinical Standards and Quality, US Renal Data System: USRDS 2000 Annual Data Report. The National Institutes of Health, National Institute of Diabetes and Digestive and Kidney Diseases, Bethesda, MD, Department of Health and Human Services, Health Care Financing Administration Bureau of Data Management and Strategy: Health Care Financing Research Report: End Stage Renal Disease, Baltimore, MD, Department of Health and Human Services, HCFA, Collins AJ, Ma JZ, Xia H, Ebben J: Trends in anemia treatment with erythropoietin usage and patient outcomes. Am J Kidney Dis 32:S133-S141, 1998 (suppl 4)