appendices with eps '01 - USRDS · rity Act of 1965 and in subsequent amendments to the Act. A person in one of these four catego-ries is eligible to apply for Medicare entitlement:

�

��

�204

204

205

207

207

208

209

209

211

215

215

216

220

223

223

222

226

226

226

229

229

230

231

232

236

238

238

��Appendix A · Analytical methods

Data sources

Data management & preparation

Database definitions

Incidence & prevalence · Chapter One & Reference Sections A & B

Patient characteristics · Chapter Two & Reference Section C

Treatment modalities · Chapter Three & Reference Section D

Clinical indicators of care · Chapter Four

Morbidity & hospitalization · Chapter Five & Reference Section E

Pediatric ESRD · Chapter Six

Transplantation · Chapter Seven & Reference Sections F & G

Survival, mortality, & causes of death · Chapter Eight & Reference Sections H & I

Cardiovascular special studies · Chapter Nine

Preventive health care measures · Chapter Ten

Provider characteristics · Chapter Eleven & Reference Section J

Economic costs of ESRD · Chapter Twelve & Reference Section K

International comparisons · Chapter Thirteen

Census population base · Reference Section L

Statistical methods

Miscellaneous

Future directions of the USRDS

Bibliography

Appendix B · ESRD network definitions

Appendix C · Glossary

Appendix D · USRDS staff & co-investigators

Appendix E · USRDS products & services

Data requests

2 0 4 � 2 0 0 1 A T L A S O F E S R D I N T H E U N I T E D S T A T E S

Appendix A · Analytical Methods

This appendix describes the USRDS database andits standardized working datasets, specializedcode definitions, and common data processingpractices. It also details the statistical methodsused in this 2001 Annual Data Report. TheResearcher’s Guide to the USRDS Database, pub-lished separately, provides additional detail aboutthe USRDS database and Standard Analysis Files.

DATA SOURCESThe USRDS maintains a stand-alone databasethat includes ESRD patient demographic anddiagnosis data, biochemical data, dialysis claims,and information on treatment history, hospital-ization events, and physician/supplier services.

REBUS/PMMIS databaseThe major source of ESRD patient informationfor the USRDS is the HCFA Renal Beneficiaryand Utilization System (REBUS), which wasadopted in 1995 as the On-Line Transaction Pro-cessing (OLTP) system from the previous Pro-gram Management and Medical InformationSystem (PMMIS) database. The REBUS/PMMISdatabase contains demographic, diagnosis, andtreatment history information for all Medicarebeneficiaries with ESRD. The database has nowalso been expanded to include non-Medicarepatients, a detailed discussion of whom is pre-sented later in this appendix.

HCFA regularly updates the REBUS/PMMISdatabase, using the Medicare Enrollment Data-base (EDB), Medicare inpatient and outpatientclaims, the United Network for Organ Sharing(UNOS) transplant database, ESRD Medical Evi-dence forms (2728) provided by the ESRD net-works, and ESRD Death Notification forms(2746) obtained from renal providers.

HCFA Medicare Enrollment Database (EDB)HCFA’s Enrollment Database is the designatedrepository of all Medicare beneficiary enrollmentand entitlement data, and provides current andhistorical information on beneficiary residence,Medicare as Secondary Payor (MSP) status, andHealth Insurance Claim/Beneficiary Identifica-tion Code (HIC/BIC) cross-referencing.

HCFA paid claims recordsInpatient transplant and outpatient dialysisclaims records are used to identify new ESRDpatients for whom no Medical Evidence form hasbeen filed. These patients, who are most likely to

be non-Medicare patients or beneficiaries whodevelop ESRD while already on Medicare becauseof age or disability, will eventually be entered intothe REBUS/PMMIS—and hence the USRDS—database through the claims records. For patientswithout Medical Evidence records these claimsare the only reliable information from which todetermine first service dates for ESRD. These paidclaims records, however, are only a supplementto—not a replacement of—other sources ofinformation on incidence and prevalence.

It is important to note that some Medicare-eligible patients may not have bills submitted toand paid by Medicare, including MSP patientscovered by private insurance, HMOs, Medicaid,or the Department of Veterans Affairs (DVA).

UNOS transplant databaseHCFA began collecting data on all Medicare kid-ney transplants in the early 1980s. In 1987, theUnited Network of Organ Sharing (UNOS) wascreated to provide a national system for allocat-ing donor organs and to maintain a scientific reg-istry on organ transplantation. UNOS also be-gan collecting data on all transplants. These twocollection efforts were consolidated in 1994, andUNOS became the single source of data on trans-plant donors and recipients.

The HCFA and UNOS transplant data files over-lap for 1987–1993, and some patients with Medi-cal Evidence forms indicating transplant as theinitial modality are not included in either file. Toresolve the conflicts among these three sources,the USRDS has adopted the following procedure:

¨ All UNOS transplants are accepted intothe database.

¨ All HCFA transplants before 1987 areaccepted.

¨ HCFA transplants from 1987 to 1993 areaccepted if there is no UNOS transplantrecord for that patient within 30 days ofthe HCFA transplant.

¨ Transplants indicated on the Medical Evi-dence forms are accepted if there is norecord of a transplant for that patientwithin 30 days of the date listed on theMedical Evidence form.

HCFA STANDARD ANALYTIC FILES (SAFs)HCFA’s SAFs contain data from final actionclaims, submitted by Medicare beneficiaries, inwhich all adjustments have been resolved. For

A P P E N D I X A � A N A L Y T I C A L M E T H O D S � 2 0 5

Part A institutional claims, the USRDS uses thefollowing data:¨ inpatient, 100% SAF¨ outpatient, 100% SAF¨ home health agency (HHA), 100% SAF¨ hospice, 100% SAF¨ skilled nursing facility (SNF), 100% SAF

For Part B physician/supplier claims:¨ physician/supplier, 100% SAF¨ durable medical equipment (DME), 100%

SAF

HCFA SAFs are updated each quarter throughJune of the next year, when the annual files arefinalized. Datasets for the current year are cre-ated six months into the year and updated quar-terly until finalized at 18 months, after which theyare not updated to include late arriving claims.Annual files are thus approximately 98% com-plete. The USRDS 2001 ADR includes all claimsup to December 31, 1999. Patient-specific demo-graphic and diagnosis information, however,includes data as recent as November 2000.

Annual Facility Survey (AFS)In addition to the HCFA ESRD databases, inde-pendent ESRD patient counts are available fromHCFA’s Annual Facility Survey, which all dialysisunits and transplant centers are required to com-plete at the end of each calendar year. The AFSreports counts of patients being treated at the endof the year, new ESRD patients starting duringthe year, and patients who died during the year.Counts of both Medicare and non-Medicare end-of-year patients are included. While AFS files donot carry patient-specific demographic and diag-nosis information, they do provide independentpatient counts used to complement the HCFApatient-specific records.

CDC SurveillanceThe Centers for Disease Control and Preventionuse their National Surveillance of Dialysis-Associated Diseases in the United States to collectinformation from dialysis facilities on patient andstaff counts, membrane types, reuse practices,water treatment methods, therapy types, vascularaccess use, antibiotic use, hepatitis vaccinationand conversion rates (for both staff and patients),and the incidence of HIV, AIDS, and tuberculosis.None of the information is patient-specific. TheUSRDS reports survey data through 1999; theCDC did not, however, conduct a survey in 1998.

DATA MANAGEMENT & PREPARATIONThe USRDS main computer system is a CompaqAlpha system consisting of one dual EV-4 (200

MHz) and two dual EV-6 (500 MHz) processorservers, with a total of 3.2 GB of RAM memoryand 1.5 terabyte (1,500 gigabytes) of disk farms,all managed by three interconnecting clusters.

We use the SASÒ database management systemand development tools as our core database tech-nology platform; this differs from the OracleRDBMS system used by the previous contractoronly in physical data allocation and management.All information in the earlier system has been in-tegrated into the new database, and its continu-ity and completeness have been maintained.

Data loading & cleaningAll data files come to the USRDS in IBM 3480cartridges/CD-ROMs with EBCDIC, ASCII, orSASÒ formats. Once loaded into the system, filesare converted into SASÒ data sets for further pro-cessing, and a series of data verification steps isexercised to ensure data quality and integrity be-fore updating the USRDS database system.

Database updatesFor this ADR, patient demographic anddiagnosis data are updated through November2000, and Medicare Part A and Part B claimsare collected through December 31, 1999.While the USRDS has generally waited 15months before reporting patient-specific datafor a given period, we intend to alleviate delaysin processing data through the Medicaresystem by reporting updated information onour website (www.usrds.org) several times eachyear. Delays have been caused by changes tothe format of the HCFA record files.

ESRD patient determinationA person is identified as having ESRD when aphysician certifies the disease on a Medical Evi-dence form (HCFA 2728), or when there is otherevidence that the person has received chronic di-alysis or a kidney transplant. Patients who expe-rience acute renal failure and are on dialysis fordays or weeks, but who subsequently recover kid-ney function, are, as much as possible, excludedfrom the database. Patients who die soon afterkidney failure without receiving dialysis treatmentare occasionally missed.

The first ESRD service date (FSD) is the singlemost important data element in the USRDS da-tabase, and each patient must, at a minimum,have a valid FSD. This date is used to determinethe incident year of each new patient and the firstyear in which the patient is counted as prevalent.The date 90 days after the FSD is used as the start-ing point for most patient survival analyses.


The FSD is derived by taking the earliest of:¨ the date of the start of dialysis for chronic

renal failure, as reported on the MedicalEvidence form,

¨ the date of a kidney transplant, as reportedon a HCFA or UNOS transplant form, aMedical Evidence form, or a hospital in-patient claim, or

¨ the date of the first Medicare dialysis claim.

Most FSDs are obtained from the Medical Evi-dence form. In the absence of this form, the dateof the first Medicare dialysis claim or transplan-tation usually supplies the FSD. In the few casesin which the date of the earliest dialysis claim isearlier than the first dialysis date reported on theMedical Evidence form, the earliest claim date isused as the FSD.

Medicare & non-Medicare (‘ZZ’) patientsBeneficiaries are enrolled in Medicare based oncriteria defined in Title XVIII of the Social Secu-rity Act of 1965 and in subsequent amendmentsto the Act. A person in one of these four catego-ries is eligible to apply for Medicare entitlement:

¨ aged 65 and over¨ disabled¨ ESRD program¨ Railroad Retirement Board (RRB)

Most ESRD patients are eligible to apply for Med-icare as their primary insurance payor. There are,however, some patients who are not immediatelyeligible for Medicare coverage because of theiremployment status and insurance benefits. Thesepatients are usually covered by Employer GroupHealth Plans (EGHPs), and must wait 30 to 33months before becoming eligible to have Medi-care as their primary payor. They are thereforenot in the EDB database during the waiting pe-riod. Some of these patients, particularly newpatients since 1995, have FSDs established byMedical Evidence forms, but have no dialysisclaims or hospitalization events in the HCFAclaims database. In the REBUS/PMMIS databasethese patients are designated ‘ZZ,’ or non-Medi-care (the REBUS/PMMIS group assigns ‘ZZ’ inthe two-character Beneficiary Identification Codefield to identify all non-Medicare ESRD patients).HCFA does not generally include these patientsin the datasets released to researchers.

The USRDS recognizes that ‘ZZ’ patients are trueESRD patients, and should therefore be includedin patient counts for incidence, prevalence, andtreatment modality. Calculations of standardizedmortality ratios (SMRs), standardized hospital-ization ratios (SHRs), and standardized trans-

plantation ratios (STRs), however, should notinclude these patients because of the small num-ber of claims available in the first 30–33 monthsafter their first ESRD service. Furthermore, it mayor may not be possible to link ‘ZZ’ patients totheir ESRD Death Notification forms (HCFA2746) or to the UNOS transplant data, and it maybe impossible to determine comorbid conditionsor Part A and Part B services. Because these dataare limited, event rates that include these patientsmust be assessed with caution.

In order to duplicate the methods used by theprevious USRDS contractor we have elected toinclude ‘ZZ’ patients in the mortality rate calcu-lations for this report. We intend to collaboratewith HCFA and other interested renal research-ers to establish, for future reports, a consistentapproach to managing the data for these patients.

Lost-to-follow-up methodologyThe USRDS draws on all available data to createa “treatment history” for each patient in the da-tabase, showing all modality events, their dura-tion, and the renal providers involved in eachpatient’s care.

Gaps frequently exist in the billing data uponwhich modality periods are based. When thesegaps occur the convention used by the USRDS isto assume that a treatment modality continuesuntil death or the next modality-determiningevent. A patient with a functioning transplant isassumed to be maintaining that transplant un-less a transplant failure or death is encounteredin the data. In the absence of death, dialysis claims,or other confirmation of a continuing modality,a dialysis modality, in contrast, is assumed to con-tinue for only 365 days from the date of the lastclaim. After this period the patient is declared lost-to-follow-up until the occurrence of a dialysisclaim or transplant event.

Because Medicare may be the secondary payorfor up to the first 30–33 months of ESRD,delaying the appearance of Medicare dialysisclaims, lost-to-follow-up categorization cannotbegin until the end of the second year after firstESRD service. This ‘first two-year rule’ isparticularly important for non-Medicare patients.Since it is now 30–33 months before these patientshave Medicare as their primary payor, somepatients may be followed for up to two years withlimited amounts of event or death data. Thesepatients contribute dialysis or transplant days tothe denominator of rate calculations, but onlyquestionable event data to the numerator. Non-Medicare patients who have been included in the


database since 1995 therefore pose a significantchallenge to the USRDS and HCFA, and methodsof tracking them are currently being explored.

A number of factors can result in a lack of dialysisdata and eventual reclassification of the patientas lost-to-follow-up:¨ The patient may have recovered renal

function and no longer have ESRD.¨ The patient may have left the country.¨ The patient’s dialysis therapy may be cov-

ered by a payor other than Medicare, orthe patient may have received a transplantnot paid for by Medicare and not reportedto UNOS.

¨ The patient may be enrolled in a Medi-care HMO, so that Medicare claims fordialysis are not generated even though thepatient is eligible for Medicare coverage.

¨ The patient’s death may not have been re-ported to the Social Security Administra-tion or to HCFA.

60-day ruleThis rule requires that a treatment modality mustcontinue for at least 60 days before it can be con-sidered a primary or switched modality. It is usedto construct a patient’s modality history, i.e. whencreating the modality sequence file.

90-day ruleThis rule defines each patient’s start date, for dataanalyses, as day 91 of ESRD. Allowing outcomesto be compared among all ESRD patients at a stableand logical point in time, this rule is used primarilywhen calculating survival rates and comparingoutcomes by modality at several points in time.Use of this rule overcomes the difficulties ofexamining data from the first three months ofESRD service, an unstable time for new patientsas renal providers try to determine the besttreatment modalities, and from center hemo-dialysis patients younger than 65 and not disabled,who cannot bill Medicare for their dialysistreatments and hospitalization until 90 days afterthe first ESRD service date (patients on peritonealdialysis or home dialysis, or with transplant as thefirst modality, can bill immediately).

DATABASE DEFINITIONSModalitiesBecause different patient modality categories areused throughout the ADR, these categories aredefined in the methods sections for each chapter.

Primary cause of renal failureInformation on the primary cause of renal failureis obtained directly from the Medicare Evidence

form. For the Annual Data Report these diseasecodes have been grouped into eight categories,with ICD-9-CM codes as follows:

¨ diabetes: 250.00 and 250.01¨ hypertension: 403.9, 440.1, and 593.81¨ glomerulonephritis: 580.0, 580.4, 582.0,

582.1, 582.9, 583.1, 583.2, 583.4, and583.81

¨ cystic kidney: 753.13, 753.14, and 753.16¨ other urologic: 223.0, 223.9, 590.0, 592.0,

592.9, and 599.6¨ other cause: all other ICD-9-CM codes

covered in the list of primary causes onthe Medical Evidence form, with the ex-ception of 799.9

¨ unknown cause: 799.9 and other ICD-9-CM codes not covered in the primarycauses on the Medical Evidence form

¨ missing cause: no ICD-9-CM code listed

Race & ethnicityInformation on patient race and ethnicity isobtained from the Medical Evidence form, theHCFA Medicare Enrollment Database, and theREBUS identification file. Because they areaddressed in separate questions on the MedicalEvidence form, racial and ethnic categories canoverlap. In both the figures and tables of this ADRwe have added expanded information onHispanic patients. Most rate calculations thatinclude this data begin with 1996, the first fullyear after the introduction of the revised MedicalEvidence Form, in which patient ethnicity is arequired field. Reference Sections A to D, however,contain Hispanic data starting in 1995, thoughthis data may be incomplete. The non-Hispaniccategory includes all patients whose ethnicity ismissing or unknown.

Because of the small number of ESRD patientsof some races, we have concentrated throughoutthe ADR on white, black, Native American, andAsian patients. As the numbers of patients ofother races increase, data on them will bepresented in the ADR.

INCIDENCE & PREVALENCE ·CHAPTER ONE & REFERENCE SECTIONS A & BIncidence is defined as the number of people in apopulation newly diagnosed with a disease in agiven time period, typically a year. Prevalence isdefined as the number of people in a populationwho have the disease at a given point in time(point prevalence) or during a given time period(period prevalence). The USRDS generally re-ports point prevalence—the type of prevalenceused primarily throughout the Annual DataReport—as of December 31, while period preva-


lence is reported for a calendar year. Annual pe-riod prevalent data thus consist both of peoplewho have the disease at the end of the year andthose who had the disease during the year anddied before the year’s end.

The USRDS treats successful transplantation asa therapy rather than as a “recovery” from ESRD.Patients with a functioning transplant arecounted as prevalent patients.

Because data are available only for patients whoseESRD therapy is reported to HCFA, patients whodie of ESRD before receiving treatment or whosetherapy is not reported to HCFA are not includedin the database. The terms incidence and preva-lence are thus qualified as incidence and preva-lence of reported ESRD. Some ESRD registries,such as the European Dialysis and Transplanta-tion Association, use the term “acceptance intoESRD therapy.” The USRDS, however, believesthat “incidence of reported ESRD therapy” ismore precise, because “acceptance” implies thatremaining patients are rejected, when in fact theymay simply not be identified as ESRD cases ormay not be reported to HCFA.

As discussed earlier, patients are classified as lost-to-follow-up if they have had ESRD for at leasttwo years but have had no reported dialysis, death,or transplant data for one year. Beginning withthe 1992 ADR, patients classified as lost-to-follow-up are not included in the point preva-lent counts; they are, however, reported separatelyin Tables B.1 and B.a of the Reference Tables.

Point prevalence is a useful measure for publichealth research, since it measures the current bur-den of the disease on the health care delivery sys-tem, and period prevalence is appropriate for costanalysis, since it indicates the total disease bur-den over the course of the year. We have chosen,however, to focus primarily on the incidence ofESRD, believing that it is the most useful mea-sure for medical and epidemiological researchthat examines disease causality and its effect ondifferent sub-populations.

The Reference Tables present parallel sets ofcounts and rates for incidence (Section A) andDecember 31st point prevalence (Section B).Section B also presents annual period preva-lent counts and counts of lost-to-follow-up pa-tients.

Incident data for Figure 1.1, a map of the oddsratios of developing ESRD, are obtained fromHCFA, while population counts are obtained

from the U.S. Census Bureau. A logistic regressionis used to compare the incidence of ESRD bylocation, with ESRD (yes or no) as the dependentvariable. Explanatory variables include incidentyear, race (white and black), gender, age (20–44,45–64, 65–74, and 75+), and location (50 statesplus the District of Columbia).

Reference Section ABecause the U.S. population figures used for thisreport (presented in Reference Section L) includeresidents only of the 50 states and the District ofColumbia, tables in this section focus on patientsfrom these areas as well. The exceptions are TablesA.a, A.9–14, and A.f–k, all of which present dataspecific to patients in Puerto Rico and the Terri-tories, or include these patients in the patientpopulation. Age is computed as of the beginningof ESRD therapy.

Reference Section BThese tables also focus on patients who are resi-dents of the 50 states and the District of Colum-bia, with the exception of Tables B.1 and B.b. Ageis calculated as of December 31 in all tables.

PATIENT CHARACTERISTICS ·CHAPTER TWO & REFERENCE SECTION CData used in both Chapter Two and ReferenceSection C are obtained from the HCFA MedicalEvidence Form (2728), which plays a central rolein the USRDS database. This form is completedat the dialysis unit for each new ESRD patienttreated at that unit, and is sent to HCFA throughthe ESRD networks. It serves to establish Medi-care eligibility for individuals who previously werenot Medicare beneficiaries, reclassify previouslyeligible Medicare beneficiaries as ESRD patients,and provide demographic and diagnostic infor-mation on all new ESRD patients regardless ofMedicare entitlement.

Before 1995 units were required to file the MedicalEvidence form only for Medicare-eligiblepatients. With the adoption of the new form in1995, however, dialysis providers are now requiredto complete the form for all new ESRD patients,regardless of Medicare eligibility status. Therevision also contains new fields for comorbidconditions, employment status, race, ethnicity,and biochemical data at the start of ESRD.

This form is the only source of information aboutthe cause of a patient’s ESRD. Because the list ofdiseases was revised for the new form, the USRDSstores the codes reported on each version so thatdetail is not lost through trying to convert oneset of codes to the other.


The data in Tables C.4 and on are restricted topatients for whom the first Medical EvidenceForm was a new form and was certified within12 months of the first service date; total patientcounts for this group are listed in Table C.3.

TREATMENT MODALITIES ·CHAPTER THREE & REFERENCE SECTION DChapter Three and the associated tables describethe treatment modalities of all known ESRDpatients, Medicare and non-Medicare, who are notclassified as lost-to-follow-up. Table D.6, however,contains data only on incident Medicare patientswho have survived at least 90 days. The 60-day ruleis used to identify a change in modality. (A detaileddiscussion of the lost-to-follow-up methodology,non-Medicare patients, and the 60-day ruleappears earlier in this appendix.)

Treatment modalities are defined here as follows:¨ center hemodialysis: hemodialysis treat-

ment received at a dialysis center¨ center self-hemodialysis: hemodialysis

administered by the patient at a dialysiscenter; a category usually combined withcenter hemodialysis

¨ home hemodialysis: hemodialysis admin-istered by the patient at home; cannot al-ways be reliably identified in the database

¨ CAPD: continuous ambulatory peritonealdialysis; usually combined with CCPD

¨ CCPD: continuous cycling peritonealdialysis; usually combined with CAPD

¨ other peritoneal dialysis: primarily inter-mittent peritoneal dialysis (IPD), a smallcategory except among very young chil-dren, and usually combined with un-known dialysis and uncertain dialysis toform an other/unknown dialysis category

¨ uncertain dialysis: a period in which thedialysis type is unknown or multiple mo-dalities occur but none last 60 days; usu-ally combined with other peritoneal dialy-sis (IPD) and unknown dialysis to forman other/unknown dialysis category

¨ unknown dialysis: a period in which thedialysis modality is not known (e.g., whendialysis sessions are performed in a hos-pital); usually combined with other peri-toneal dialysis (IPD) and uncertain dialy-sis to form an other/unknown dialysiscategory

¨ renal transplantation: a functioning graftfrom either a living donor (a blood rela-tive or other living person) or a cadavericdonor

¨ death: a category not appearing in theyear-end modality tables, which report

only living patients, but used as an out-come, i.e., in tables showing living patientsfollowed for a period of time for theirmodality treatment history

The tables in Reference Section D are divided intothree sections. The first section, Tables D.1–5 andD.10, provides counts and percentages, by demo-graphics and treatment modality, of prevalentpatients alive at the end of each year. Because thesetables include both Medicare and non-Medicarepatients, counts and percentages in the catego-ries of unknown age, gender, race, primary causeof renal failure, network, and state are significantlyhigher. The totals by year are also higher than thepoint prevalent counts in Reference Section B,which include only U.S. residents and droppatients with missing birth dates. Age is com-puted as of December 31.

The second section, Table D.6, shows modalityat 90 days and two years after first service for allincident Medicare patients beginning renal re-placement therapy from 1995–1997. The 90-dayrule is used to exclude patients who died duringthe first 90 days of ESRD, and age is computed asof the date of first ESRD service.

The third section, Tables D.7–8, presents countsof prevalent patients alive at the end of each yearby ESRD exposure time and modality. Table D.7shows counts by the number of years the patienthas had ESRD, while Table D.8 shows counts bythe number of years on the end-of-year treatmentmodality. For the duration of ESRD exposure,zero should be read as less than one year, one asat least one full year but less than two, and so on.

CLINICAL INDICATORS OF CARE · CHAPTER FOURData underlying the figures in this chapter areobtained from several sources. Erythropoietin(EPO) dose information and hematocrit valuesin Figures 4.1–15 are obtained from EPO claimsdata, while in Figures 4.16–22 Part B physician/supplier claims data supply the CPT codes indi-cating the insertion of temporary and permanentcentral venous accesses and simple fistulas. Theinformation used in Figures 4.23–24 and 4.26–27 is obtained from the HCFA 2000 ESRD Clini-cal Performance Measures Project (formerly theESRD Core Indicators Project). Data on urea re-duction ratios (URRs) in Figure 4.25 come fromPart A institutional outpatient claims. All figuresinclude data from Medicare patients only.

Figure 4.1 shows the mean entry-period hema-tocrit and first- and second-year death rates inincident hemodialysis patients from 1990 to 1998.


Patients are classified by incident year, defined asthe year of the first ESRD service date. Those in-cluded in this graph survive a full 90 days plus asix-month entry period, in which they have atleast one EPO claim. The mean entry-periodhematocrit is calculated for each patient, and thepercentages of patients with mean hematocritswithin the following ranges are graphed for eachincident year: <30%, 30–<33%, 33–<36%, and³36%. The adjusted first- and second-year deathrates are then calculated from the Cox model,adjusted by age, gender, race, and primary diag-nosis. Methods used follow the interval death ratemethods discussed in the statistical methods sec-tion later in this appendix.

Figure 4.2 displays mean hematocrit by state. Thedata include 1999 prevalent hemodialysis patientswith at least one EPO claim during the year and£20 EPO administrations per month. Time at riskbegins on January 1 (prevalent patients) or day91 of ESRD (incident patients) and is censoredat the earliest of modality change, loss-to-follow-up, death, or December 31, 1999. Each patient’syearly mean hematocrit is calculated from claimsduring the time at risk, and the average of thesevalues is computed for each state.

Figure 4.3 shows the mean hematocrit, bymonth (1996 through May 2000), for preva-lent ESRD patients with EPO claims, alongwith the monthly dose per week for prevalenthemodialysis and peritoneal dialysis patientswith EPO claims and £20 EPO administrationsper month. Figure 4.4 illustrates the distribu-tion of patients by mean hematocrit group. Toobtain the data for both figures a monthlymean hematocrit is computed for each patient,and the mean dose per week is calculated asthe total units in the month divided by the to-tal weeks. This mean EPO dose per week, how-ever, does not take into consideration the ac-tual number of weeks EPO is administered, forthere may be gaps in administrations frommissed or held doses. These unweighted aver-ages may thus overestimate the weeks in thedenominator and underestimate the true EPOdose per week. In order to give more weight topatients consistently receiving EPO, weightedaverages across patients are calculated, usingthe number of monthly administrations as theweight. A monthly mean hematocrit andweighted mean dose per week are then calcu-lated across all patients. For hemodialysis pa-tients a weight of 1 is thus assigned to thosewith 11–14 administrations, a weight of 10/11 to10 or 15, 9/11 to 9 or 16, etc.; for peritoneal dialy-sis patients, a weight of 1 is assigned to those

with 4–7 administrations, a weight of 3/4 to 3or 8, 1/2 to 2 or 9, etc.

Figure 4.5 illustrates the average hematocrit forprevalent hemodialysis and peritoneal dialysispatients with EPO claims. Each patient’s yearlymean hematocrit is calculated from claims datedfrom January 1 or day 91 of ESRD until a modal-ity change or the end of the year, whichever isearliest, and the average of these values for eachyear and dialysis group is computed.

Figures 4.6–9 present data by race and modalityfor prevalent hemodialysis and peritoneal dialy-sis patients with EPO claims. For each year, EPOdata is obtained from claims dated from January1 or day 91 of ESRD until a modality change orthe end of the year, whichever is earliest. Figure4.9 includes only those patients who have bodyweight data on the HCFA Medical EvidenceForm, 1995 to 1999.

In Figures 4.8 and 4.9 the mean EPO dose perweek, weighted by the average monthly admin-istrations, is calculated with a method similar tothat used in Figure 4.3. However, instead of be-ing calculated on a monthly basis, the mean doseper week in Figures 4.8 and 4.9 is calculated onan annual basis, in which it is weighted by theaverage—rather than actual—number ofmonthly EPO administrations. The averagenumber of monthly EPO administrations, cal-culated as the total administrations divided bythe total time at risk in months, is therefore usedas the weight. For hemodialysis patients, a fullweight of 1 is given to patients with an average of>11–14 administrations per month. Linearly de-creasing weights are given to patients with 11 orfewer or with greater than 14 administrations permonth. The following weights are assigned tohemodialysis patients with the given number ofaverage monthly EPO administrations: 11/12 for>10–11 or >14–15; 10/12 for >9–10 or >15–16; 9/12

for >8–9 or >16–17; 8/12 for >7–8 or >17–18; 7/12

for >6–7 or >18–19; 6/12 for >5–6 or >19–20; 5/12

for >4–5; 4/12 for >3–4; 3/12 for >2–3; 2/12 for >1–2;and 1/12 for >0–1. For peritoneal dialysis patients,a weight of 1 is assigned to patients with an aver-age of >4–7 administrations per month, and thefollowing weights are assigned to the givenmonthly administrations: 4/5 for >3–4 or >7–8;3/5 for >2–3 or >8–9; 2/5 for >1–2 or >9–10; and1/5 for >0–1 or >10–20. This method is also usedin Figures 4.11 and 4.14.

Mean hematocrit (Figures 4.10 and 4.13), meanEPO dose per week (Figures 4.11 and 4.14), andthe percent of patients who meet the DOQI tar-


get hematocrit of ³33% (Figures 4.12 and 4.15)are presented for 1999 period prevalent hemo-dialysis and peritoneal dialysis patients with EPOclaims. Multiple hematocrit measurements arecondensed into one mean value per patient, andthese values are used to calculate the mean hema-tocrit across patients for each HSA as well as thepercentage of patients with mean hematocrit³33%. Mean EPO dose per week is computedacross patients on an HSA level and weighted bythe average number of monthly administrations,following the methods used for Figures 4.8–9.

Figure 4.16 displays catheter days per patient yearat risk and days per insertion for 1996–1999period prevalent hemodialysis patients, whileFigures 4.17–18 show these rates on an HSA levelfor 1999. Figures 4.19–21 show HSA-level mapsof temporary and permanent central venouscatheter insertion rates and simple fistula creationrates per 1,000 patient years at risk, and Figure4.22 illustrates these annual rates for 1996 to 1999.

Using the same method applied in the hospital-ization analyses, hemodialysis patients failing toreach a certain level of Medicare paid dialysis billsare excluded from Figures 4.16–22, as are thoseclassified in the enrollment database as MSP any-time during the follow-up period. Part B physi-cian/supplier claims data provide the CPT codesfor these insertion rates; specific codes are listedin the figure captions. Duplicate occurrences ofthe same CPT code with the same first and lastexpense date on multiple lines of a claim arecounted as one insertion. Time at risk begins onJanuary 1 (prevalent patients) or day 91 of ESRD(incident patients) and is censored at the earliestof modality change, loss-to-follow-up, death, orDecember 31, 1999.

In the calculation of temporary and permanentcentral venous catheter insertion rates and cath-eter day rates (Figures 4.16–20 and 4.22), addi-tional methods are used to exclude catheters in-serted for purposes other than dialysis. A CPTcode of 36533, 36489, or 36491 is included onlyif it is associated with either a line-level diagnosiscode or a claim-level principal diagnosis codeamong the following ICD-9-CM codes relatedto dialysis or renal failure: 250, 403, 580–589, 593,996.1, 996.62, 996.73, V45.1, and V56. Addition-ally, Part B physician/supplier and durable medi-cal equipment claims are searched for chemo-therapy (CPT codes 96408, 96410, and 96412)and parenteral nutrition (CPT codes B4164–B5200, B9004, B9006, and B9999) claims. Patientswith at least one of these codes during a year areexcluded for that year.

Catheter days during the time at risk are countedby summing the days from a catheter insertion(CPT code 36533) until a catheter removal (CPTcode 36535). Since insertions and removals oc-cur on claims on separate lines with different ex-pense dates, it is possible for two insertions ortwo removals to occur consecutively, in whichcase the days are counted from the first insertionor until the last removal. When the first codeduring the time at risk indicates a removal, cath-eter days are counted from the beginning of thetime at risk until the removal. Similarly, in thecase of an insertion as the last code of the time atrisk, catheter days are counted from the insertionuntil the end of the risk period. An insertion andremoval occurring on the same day are countedas one catheter day. Rates of catheter days per yearat risk are then calculated by dividing the totalcatheter days during the time at risk by the totalyears at risk, and the number of days per inser-tion is computed by dividing the total days bythe number of insertions during the time at risk.

Figure 4.25 illustrates the percent of 1999 preva-lent hemodialysis patients with a median ureareduction ratio of ³65%, the target set by the NKFDialysis Outcomes Quality Initiative (DOQI).The range of each patient’s URR is obtained fromthe “G” modifier attached to CPT code 90999with revenue codes 821 or 825, and the medianrange includes all URR values from January 1 orday 91 of ESRD until the end of the year. When apatient has an even number of URR ranges, thetwo middle values are each given a weight of 0.5.

MORBIDITY & HOSPITALIZATION ·CHAPTER FIVE & REFERENCE SECTION EChapter FiveMethodologies used to create the figures in thischapter generally echo those used to create thetables in Reference Section E (described below).Inclusion and exclusion criteria are the same, asare the methods for computing hospitalizationrates. Part A inpatient institutional claims are usedfor the analyses unless otherwise specified, andthe methodologies for excluding MSP patientsare applied here as well, as detailed in the discus-sion of Section E.

One exception is the practice in the referencetables of combining hospitalizations that oc-cur with no days between discharge and thefollowing admission into one hospitalizationfrom the first admission date to the last dis-charge date. For Chapter Five, the only con-secutive hospitalizations combined into oneare those that overlap. In this case the princi-pal diagnosis and procedure codes are retained


from the first of the two overlapping hospital-izations, with the combined hospitalizationextending from the first admission date to thelast discharge date. An admission that occursthe same day or the day after a discharge is des-ignated as a separate hospitalization with itsown diagnosis and procedure codes.

New to this year’s ADR, Figures 5.3–5, 5.11, and5.18–20 present data for Hispanic versus non-Hispanic patients according to the ethnicity clas-sification on the HCFA Medical Evidence form.Patients whose ethnicity is reported as unknown,missing, or other than Hispanic are included inthe non-Hispanic category.

Data for Figures 5.1–5, and for related figures, arecalculated using different methods. Data onhospital days per patient year at risk include onlydays within the analysis period, while data onhospital days per admission include all days forhospitalizations in which the admissions occurwithin the analysis period, even days occurringafter the period has ended. The number of daysper admission in Figures 5.3–6 thus representsthe mean length of stay per admission forhospitalizations beginning within the time at riskfor the given year. The number of days per patientyear at risk (Figures 5.1–5, 5.8, 5.15, and 5.18–20), however, includes only those hospital daysthat fall within the time at risk, regardless of theadmission date.

Figures 5.6–7 display HSA-level hospital admis-sions per patient year at risk and days per admis-sion for 1999 prevalent ESRD patients. Maps arepresented separately for whites, non-whites, andall ESRD patients; the non-white group containspatients with race listed as missing, unknown, orother than white.

Figure 5.8 presents HSA-level data of the numberof hospital days per year at risk for prevalentESRD patients. While time at risk is censored hereat death or the end of the year for the all-ESRD,diabetic, and non-diabetic patients, hemodialysisand peritoneal dialysis patients are also censoredat three days prior to transplant in order to excludetransplant-related hospital days. Transplantpatients are also censored at three years followingthe date of transplant, since their Medicareeligibility may be lost and their hospitalizationdata may be incomplete. The diabetic groupconsists of all ESRD patients for whom diabetesis stated as the cause of disease, while the non-diabetic group contains all ESRD patients with acause of disease that is unknown, missing, orother than diabetes.

Data from 1997 to 1999 are combined in Figures5.9–11 and 5.15, which illustrate the mean num-ber of hospital admissions and average length ofstay per year at risk by gender (or, in 5.11, byethnicity) and modality in combination with age,race, or diabetic status. Age is determined onJanuary 1 of the year, and patients with a cause ofdisease that is unknown, missing, or other thandiabetes are classified as non-diabetic. Patientswith missing age, gender, or race, as well as pa-tients with race other than white, black, NativeAmerican, or Asian, are excluded.

Figures 5.12–14 show the number of hospital-izations per patient year at risk for incident andprevalent hemodialysis, peritoneal dialysis, andtransplant patients at various vintages, or timesfollowing the first ESRD service date. Vintage isdefined as the time from the first ESRD servicedate until January 1 of the year for prevalentpatients, or, for incident patients, until day 91 ofESRD. Data is presented by year for 1991 to 1999,with the year representing the year following thefirst service date. A patient with a first service dateof April 5, 1990, for example, is included in the<1 year group for 1991, in the 1–<2 years groupin 1992, and in the 2–<3 years group in 1993.Time at risk is defined as it is in Figure 5.8, withtransplant patients censored at three years follow-ing the date of their most recent transplant. (Atransplant patient is included if he or she has beenan ESRD patient for longer than three years, aslong as the most recent transplant was less thanthree years ago.)

Data on the frequency of principal proceduresand diagnoses (Figures 5.16–17) and on relatedhospital days (Figures 5.18–20) are obtained fromPart A inpatient and outpatient institutionalclaims. Patients with missing values for genderor race are excluded. The time at risk for eachprocedure is censored at the end of the year, death,or three days prior to transplant. As in the totaladmission rates presented in the hospitalizationreference tables, inpatient rates are calculated bysubtracting the days spent in the hospital for eachprocedure or diagnosis from the total time at riskfor admission for that procedure or diagnosis.

Most of the principal procedure and diagnosiscodes used in Figures 5.16–20 are listed in thefigure captions. Infection (overall) is indicatedby principal ICD-9-CM codes of 001–139,254.1, 320–326, 331.81, 372–372.39, 382–382.4,383.0, 386.33, 386.35, 388.60, 390–393, 421–422, 460–466, 472–474.0, 475–477.9, 478.22–478.24, 478.29, 480–491, 494, 510–511, 513.0,518.6, 522.5, 522.7, 527.3, 528.3, 540–542, 566–


567.9, 569.5, 572–572.2, 573.1–573.3, 575–575.12, 590–590.9, 595–595.4, 597–597.89,599.0, 601–601.9, 604–604.9, 607.1, 611.0, 614–616.1, 616.3–616.4, 616.8, 670, 680–686.9,706.0, 711–711.9, 730–730.3, 730.8–730.9,790.7–790.8, 996.6–996.69, 997.62, 998.5,999.3, V01–V069, V08, and V09.

Figures 5.21–30 show the relative risk ofhospitalization by URR and hematocrit. Thestudy includes adult (>19 years of age) 1998incident hemodialysis patients, for whom theincident date is defined as the first ESRD servicedate plus 90 days. Included patients survive thefirst 90 days plus a full six-month entry period,and have at least four EPO claims and three URRmeasurements during the entry period. Patientswith a bridge hospitalization spanning the startof the follow-up period are excluded. The rangeof each patient’s URR is obtained from the “G”modifier attached to CPT code 90999 withrevenue codes 821 or 825. For each patient themedian URR of the last three entry-period URRvalues is selected, and the mean entry-periodhematocrit is computed. Patients are followedfrom the end of the entry period until the earliestof death, first hospitalization, modality change,loss-to-follow-up, or December 31, 1999.

In Figures 5.21–22, time to first all-causehospitalization is assessed using a Coxproportional hazards regression model stratifiedby primary cause of renal failure, and includingcovariates of age, gender, comorbidity, diseaseseverity (number of hospitalizations, bloodtransfusions, and vascular access procedures inthe entry period), hematocrit (<30%, 30–<33%,33–<36%, and ³36%), and URR (<60%, 60–<65%, 65–<70%, 70–<75%, and ³75%).Separate models are run for whites and blacks,excluding patients of other races. Patients withhematocrits of 33–<36% and URRs of 65–<70% are used as the reference, and the relativerisks and 95% confidence intervals are displayedin Figures 5.21–22. Similar methods are used inFigures 5.23–30; in Figures 5.23–24, however,separate models are run for diabetic and non-diabetic patients, stratified by race. Additionally,in Figures 5.25–26, the model is stratified byprimary cause of renal failure, and race isincluded as a covariate. Also following thesemethods, Table 5.1 shows the results of separateCox proportional hazards regression models forwhite and black patients, including primarycause of renal failure as a covariate, while Table5.2 shows the results of separate models fordiabetic and non-diabetic patients, includingrace as a covariate.

The methods of Tables 5.1–2 and Figures 5.21–24 are repeated in Tables 8.2–3 and Figures 8.18–21 for all-cause death. The methods of Figures5.25–26 are repeated in Figures 5.27–30 for car-diovascular and infectious hospitalization, andagain in Figures 8.22–27 for all-cause, cardiac, andinfectious death. Figures 8.22–27, however, do notexclude patients with a bridge hospitalization, andpatients are followed from the end of the entryperiod until the earliest of death, modality change,loss-to-follow-up, or December 31, 1999. Infec-tious hospitalization is defined by the principalICD-9-CM codes used in Figures 5.16–17 and5.20, while cardiovascular hospitalization is de-fined by the following principal ICD-9-CMcodes: 276.6, 394–398.99, 401–405, 410–438, and440–459. Cardiac death is defined by a primarycause of death code of 01, 02, 23, or 25–31 on theESRD death notification form, while infectiousdeath is defined by a primary cause of death codeof 10–13, 49–60, 64–65, or 74.

Reference Section EBecause hospitalization data may be incompletefor non-Medicare patients, the analyses in thissection include Medicare patients only. Hospi-talization data is obtained from Part A institu-tional inpatient claims, and Table E.27 also in-cludes REBUS hospitalization data through June1998. REBUS inpatient hospitalization data fromJuly 1998 to December 1999 is not included inthis table because it is currently (as of March 2001)unavailable from HCFA.

Tables E.1–26 include dialysis and transplantpatients who have been on their modality for atleast 60 days, who have reached day 91 of ESRDby the end of the year, and who are residents ofthe 50 states, the District of Columbia, PuertoRico, or the Territories. Patients with AIDS as aprimary or secondary cause of death, patientswith missing values for age, gender, or race, andpatients of races that are unknown or other thanwhite, black, Native American, or Asian areexcluded. Age is classified on January 1 of eachyear. Patients are also classified according to theprimary cause of ESRD, in which the “other”category includes patients with missing data orcauses other than diabetes, hypertension, orglomerulonephritis.

Patients are classified by modality at the begin-ning of the year using the following categories:

¨ all-dialysis: patients on hemodialysis,CAPD/CCPD, or dialysis of an un-known type, as well as patients who havenot been on one modality for the previ-ous 60 days


¨ hemodialysis: patients who have been onhemodialysis for at least 60 days at the startof the period

¨ CAPD/CCPD: patients who have been onCAPD/CCPD for at least 60 days at thestart of the period at risk

¨ all-ESRD: all patients

To limit the contribution of patient years at riskfrom patients who are classified as MSP, and whotherefore have incomplete hospitalization data,dialysis patients failing to reach a certain level ofMedicare paid dialysis bills are excluded fromTables E.1–26. Dialysis patient start dates (January1 of the year for prevalent patients and day 91 ofESRD for incident patients) must fall betweenstart and end dates based on Medicare paiddialysis claims, as follows:

¨ start date: the first day of the first monthin which there is at least $675 of Medicarepaid dialysis claims

¨ end date: the end of a three-month pe-riod in which there is less than $675 of paidclaims in each month

If a patient’s start date does not fall between thesedates, he or she is excluded from the analysis forthat year. This method is similar to that used inthe economic analysis section, except that herethe paid claims dates are analyzed only for thedialysis patient start date. The dialysis patient enddate remains the earliest of death, three days priorto transplant, and December 31 of the year.

MSP patients are also excluded from this datasetthrough use of an additional filter that usesinformation from the enrollment database todetermine MSP status. If this database indicatesMSP status anytime during the time at risk, thepatient is excluded for the year.

For patients in the all-dialysis, hemodialysis, andperitoneal dialysis categories, the period at riskfor all hospitalization analyses is from January 1or day 91 of ESRD until the earliest of death, threedays prior to transplant, or December 31.Modality change is considered a censoring eventonly in the case of a change from dialysis totransplant. For dialysis patients in the all-ESRDcategory, in contrast, the analysis period forhospitalization is censored only at death orDecember 31 of the year; modality change is notused as a censoring event. For transplant patientsin the all-ESRD and transplant categories, theanalysis period is censored at the earliest of death,three years after the transplant date, or December31 of the year. The censoring of transplantpatients at three years following the transplant is

new to this year’s ADR, and necessary becauseMedicare eligibility may be lost andhospitalization data may be incomplete for thesepatients.

In the case of a hospitalization that begins priorto January 1 or day 91 of ESRD and continuesinto the analysis year, the time at risk for firstadmission begins the day after discharge fromthis bridge hospitalization. Patients with abridge hospitalization that spans the entireanalysis period are excluded from the firstadmission rates.

Time at risk is calculated differently for length ofstay and total admissions. Since a hospitalizedpatient remains at risk for additional hospitaldays, rates for hospital days include hospital daysin the time at risk. Since a currently hospitalizedpatient is not, however, at risk for new admissions,hospital days for each year are subtracted fromthe time at risk for total admissions. In the caseof hospitalizations in which admission occurs thesame day as discharge, zero days are subtractedfrom the time at risk for total admissions. Whenbridge hospitalizations span the start of theanalysis period, only the days within the periodare subtracted from the time at risk for totaladmissions.

All admissions and hospital stay days during theanalysis period are included, respectively, in thetotal admissions and length of stay for each year.An admission for a hospitalization that occursbefore and spans the start of the analysis periodis excluded from the total admissions for thatanalysis period, and only the hospitalization dayswithin the period are counted in the total daysfor length of stay rates. The minimum length ofstay is one day, and hospitalizations withadmission and discharge on the same day, as wellas hospitalizations with discharge the day afteradmission, are both counted as one day.

In Tables E.1–26, overlapping hospitalizationsand hospitalizations that occur with no daysbetween discharge and the following admissionare combined into one hospitalization that spansthe first admission date to the last discharge date.This follows the methodology of previous AnnualData Reports, allowing for comparison of thereference tables across years.

Table E.27 in contrast, includes all hospitalizationsin the total discharges reported, and nooverlapping or adjacent hospitalizations arecombined. These tables present total hospitaldischarges by diagnostic related groups (DRGs),


and no exclusions are made for patients who diedof AIDS or for MSP status. Total discharges arepresented by modality group and the year ofdischarge. For each year the total discharges arecounted from January 1 or the first ESRD servicedate until the end of the period at risk, as definedpreviously. In this case, however, the period at riskfor transplant patients in the transplant and all-ESRD groups is not censored at three yearsfollowing the date of transplant. Inpatient REBUSdata (available only through June 1998) arecombined with Part A institutional inpatientclaims data, and duplicate observations from bothsources with identical hospitalization start dates,end dates, and DRG codes are omitted.

The methodology used in Tables E.1–5 for com-puting first admission rates uses a generalizedmixed model. Smoothed rates are used to calcu-late the expected number of first hospitalizations,a number then used to obtain the standardizedhospitalization ratios (SHRs) in Table E.6. Thesemethods are described later in this Appendix.

The methods used to compute total admissionsand days hospitalized are the same as those usedin prior Annual Data Reports. As in the 2000ADR, the total admission rate is expressed per1,000 patient years at risk, while the rate of hos-pital days is given per patient year at risk. Datafrom 1997 to 1999 are pooled to increase stabil-ity, but follow-up is for single calendar year peri-ods using cohorts of patients alive at the begin-ning of each year. The number of hospitaladmissions and days, and the number of years atrisk for each event, are computed separately foreach year and summed over the three years; ratesare then computed by dividing the total admis-sions or days by the total time at risk. A patientwho is alive at the beginning of 1997, dies in 1999,and has two hospitalizations each year, for ex-ample, will contribute two and a fraction years atrisk and six admissions.

PEDIATRIC ESRD · CHAPTER SIXInformation on pediatric patients is a subset ofthe ESRD patient data used throughout the ADR;methods used to create most figures in this chap-ter are therefore the same as those described inthe related chapter discussions.

The hospitalization data presented in Figures6.15–17 contain information from Part Ainstitutional inpatient claims, and exclude non-Medicare patients. As in the methods used forChapter Five, two overlapping hospitalizationsare combined into one, and an admission thatoccurs the same day or the day after a discharge

is counted as a separate hospitalization. Patientswith AIDS as a primary or secondary cause ofdeath are excluded, as are patients classified asMSP according to paid dialysis claims or theenrollment database. Residents of the 50 states,the District of Columbia, Puerto Rico, and theTerritories are included. Age is classified onJanuary 1 of the year, and patients with missingage or gender information are excluded. PriorESRD time is calculated as the time from the firstESRD service date until the first of the year forprevalent patients, or day 91 of ESRD for incidentpatients. Principal ICD-9-CM diagnosis codesused for overall infection are listed in thediscussion of Figures 5.16–20.

TRANSPLANTATION ·CHAPTER SEVEN & REFERENCE SECTIONS F & GChapter SevenIn addition to the analyses conducted for the ref-erence tables (discussed below), several additionalmethods are used for the figures in this chapter.All figures which present graft or patient survivalestimates exclude non-Medicare patients (due tolack of follow-up information) and patients withfirst ESRD service dates prior to 1977.

Figure 7.1 presents organ donation rates. Thenumerators include all transplant recipientsyounger than 65, with known gender, and whoare white, black, Native American, or Asian. De-nominators are obtained from the U.S. Census.Rates are calculated as the number of donatedkidneys (excluding discarded organs) divided bythe population, and multiplied by one million toyield donations per million population. Theserates are mapped by HSA in Figures 7.4 and 7.5.

Figure 7.9 presents trends, by race, in graft andpatient survival rates after first transplant, alongwith estimates of conditional graft and patient half-life. Survival rates are adjusted for age, gender, andprimary diagnosis, and standardized to 1997. Half-life estimates are conditional on surviving one yearpost-transplant. The estimated median half- life iscalculated by multiplying the estimated meansurvival by log(2). The estimated mean survival iscalculated as the sum of the survival time dividedby the number of observed graft failures (includingdeath) and patient deaths, respectively.

Figure 7.10 presents trends in the annual rate ofgraft loss, calculated from a Cox model andadjusted for age, gender, race, and primary diag-nosis. Patient follow-up is censored at December31, 1999. In estimating graft survival, death is con-sidered a graft failure. Graft survival probabilitiesare standardized to the 1997 patient population,


and the estimated percent graft failure is calcu-lated as one minus the estimated graft survivalprobability for each year.

Reference Section FTransplant counts are presented in Tables F.1–16. All known transplant events are includedunless specified in the footnote, and all countsinclude non-Medicare patients.

Calculations of transplant rates per 100 patientyears on dialysis begin in Table F.17, and includeonly patients reaching day 91 of ESRD service.All hemodialysis patients, peritoneal dialysis(CAPD/CCPD) patients, and patients on an un-known form of dialysis are included, as are allnon-Medicare patients. Patients who died ofAIDS or whose age is unknown at transplant areexcluded. A patient’s dialysis days are countedfrom the beginning of the specified year, or dayone of ESRD dialysis therapy if treatment beginsmid-year, until the first of transplant, death, orthe end of the year. Patients lost-to-follow-up ina given year are not censored at the lost-to-follow-up date, but are followed until the end of the cal-endar year. Transplant rates are calculated as thenumber of transplant events divided by the totalnumber of dialysis patient years for each year.

Table F.19, first transplant rates per 1,000 patient-years at risk, is calculated using a generalizedmixed model to stabilize the rates (this model isdetailed later in this appendix).

Table F.24 displays standardized first transplantratios by state and territory for 1997–1999. Astate’s observed first transplant rate is comparedto the rate expected from national rates forpatients with similar characteristics, with the 1999prevalent cohort (Table F.19) used as a reference.The standardized first transplant ratio is calcu-lated as the ratio of the observed number of firsttransplants in the state to the expected number.

Reference Section GThis section presents graft survival probabilitiesfor various demographic groups and lengths offollow-up. Patients are followed from the trans-plant date to graft failure, death, or the end of thefollow-up period (December 31, 1999); death inthis analysis is considered a graft failure. Becausea minimum of one year of follow-up is needed,1998 is the most recent year reported.

To produce a standard patient cohort for thegraft survival analyses, patients with unknownage, gender, or race are omitted. Unknown ageis defined as a missing age at transplant, or an

age that was calculated to be less than zero orgreater than or equal to 100. Patients are alsoexcluded if their first ESRD service date is priorto 1977. Non-Medicare patients are excludedfrom all tables in this section due to the lack offollow-up information; the renal transplantcounts presented here differ, therefore, fromthose in Section F.

Unadjusted survival probabilities are estimatedwith the Kaplan-Meier method and Greenwood’sformula, while the Cox model is used for adjustedprobabilities. Probabilities are adjusted for age,gender, race, and primary diagnosis, standardizedto 1997 patient characteristics, and expressed aspercentages from 0 to 100.

SURVIVAL, MORTALITY, & CAUSES OF DEATH ·CHAPTER EIGHT & REFERENCE SECTIONS H & IChapter EightFigures in this chapter are created using the samemethodologies and patient populations used inthe related Reference Tables (described below);methods unique to the figures are discussed here.

Figure 8.1 presents adjusted incident and pointprevalent rates of reported ESRD, and adjustedfirst-year death rates for all incident ESRDpatients. The adjustment methods and death ratecalculations used here and throughout the chap-ter are described in the Statistical Methods sec-tion of this appendix.

Figure 8.2 presents adjusted first-year death ratesby state for 1998 incident dialysis patients.Residents of the 50 states, the District ofColumbia, Puerto Rico, and the Territories areincluded, as are all non-Medicare patients.Patients whose gender or date of birth are missingin the database, or whose age is listed as greaterthan 120, are excluded. Patients are followed fromthe first ESRD service date plus 90 days, and arecensored at transplant, loss-to follow-up, orDecember 31, 1999. The event is death duringthe one-year follow-up period. Covariates in themodel include age, race, gender, and primarydiagnosis of diabetes, and the stratum is state. Thereference for each state is the populationdistribution of the 1998 incident dialysis patients.

Figure 8.3 presents adjusted first-year death ratesby state for patients whose first transplants failand who return to dialysis in a calendar year. The1995 dataset, for example, includes only thosepatients whose first transplant was in 1995, andwho returned to dialysis that same year becausethe transplant failed. Patients are followed for oneyear from the date of their return to dialysis, and


are censored at a second transplant or a loss-to-follow-up. Residents of the 50 states, the Districtof Columbia, Puerto Rico, and the Territories areincluded, as are all non-Medicare patients, whilepatients whose gender or date of birth are missing,or whose age is listed as greater than 120, areexcluded. Covariates include age, race, gender,primary diagnosis of diabetes, and prior time onESRD. The stratum in the model is state, and thereference is the population distribution of allincluded patients from 1995 to 1998.

Table 8.1 reports expected remaining lifetimes byage, gender, race, and modality. These are calcu-lated after excluding deaths due to AIDS, acci-dents (“accidents unrelated to treatment” on theESRD Death Notification), and illegal drugs(“drug overdose (street drugs)”), so the lifetimesreported correspond to hypothetical populationsin which these causes of death do not occur.

The expected remaining lifetime for a patientgroup is the average of the life expectancies forthe patients within that group. Some patients inthe cohort will live longer than, and some lessthan, the average. Although the average cannotbe known until all patients in the cohort havedied, the expected remaining lifetime can beprojected by assuming that patients in the cohortwill die at the same rates as those observed amonggroups of recently prevalent ESRD patients.

For a subgroup of ESRD patients of a particularage, the expected remaining lifetime iscalculated using a survival function, in turncalculated using the observed death rates. Letr(A) denote the death rate for a five-year agegroup, with A identifying one of the listed age

ranges. Death rates for successive age intervals,r(A), are plotted versus age, A, and the areaunder the curve up through age A is denotedby R(A). The survival function, S(A), at age Ais the fraction of patients that would surviveto age A in a hypothetical cohort subjected tothese death rates throughout their lifetimes.The survival function at age A is related to thedeath rates by the equation S(A) = exp(-R(A)),where “exp” denotes the exponential function.Among patients alive at age A, the fraction whosurvive for X more years is then SA(X) =S(A+X) / S(A). For a given starting age, theexpected remaining lifetime is then equal to thearea under the curve of SA(X) plotted versusX. Because few patients live beyond 100, thisarea is truncated at the upper age limit A + X =100.

Figures 8.16 and 8.17, each broken down by dia-betic status, show the association of mortality andhematocrit level for the 1994–1997 cohorts ofincident hemodialysis and CAPD/CCPDpatients. The cohorts analyzed for these graphsare defined by the year (1994–1997) of first ESRDservice. Excluded from the analysis are patientsliving in the Territories, incident patients who didnot survive 90 days after the onset of ESRD, andpatients with fewer than four hematocrit claims.All causes of death are included. For each inci-dent cohort, data on patient characteristics,comorbidity, and disease severity are collectedduring the first six months (the entry period).Demographic characteristics include age, gender,and race, while comorbid conditions consist ofatherosclerotic heart disease, congestive heartfailure, peripheral vascular disease, cerebrovascu-lar accident/transient ischemic attack, other car-

Table a.1Collapsed cause of deathcategoriesused for Figures 8.14–15

Acute myocardial infarction Cardiac arrest, cause unknown Cardiac, other

Atherosclerotic heart disease Cardiac arrhythmia Cardiomyopathy Pericarditis, including cardiac

tamponadePulmonary edema due to exogenous

fluidValvular heart disease

Cerebrovascular (CVD, CVA/TIA)Cerebrovascular accident Ischemic brain damage

Infection AIDS Fungal peritonitis Hepatitis B Infection, other Other viral hepatitis Pulmonary infection, bacterial Pulmonary infection, fungal Pulmonary infection, other

Septicemia, peritonitis Septicemia, PVD/gangrene Septicemia, vascular access Septicemia, other Tuberculosis Viral infection, CMV Viral infection, other

Malignancy Malignant disease, history of

immunosuppressive treatmentMalignant disease, other

Other known causes Accident related to treatment Accident unrelated to treatment Air embolism Bone marrow depression Cachexia Chronic obstr. pulmonary disease Cirrhosis Complications of surgery Dementia Diabetic coma, hypo/hyperglycemia Drug overdose

Drug overdose, street drugs GI hemorrhage Hemorrhage, dialysis circuit Hemorrhage, ruptured vascular access Hemorrhage, surgery Hemorrhage, transplant site Hemorrhage, vascular access Hyperkalemia Liver failure, cause unknown Liver-drug toxicity Mesenteric infarction/ischemic bowel Other hemorrhage Other identified cause Pancreatitis Perforation of bowel Perforation of peptic ulcer Polycystic liver disease Pulmonary embolus Seizures Suicide

UnknownIncludes causes of death not recorded or no Death Notification form


diac disease, cancer, chronic obstructive pulmo-nary disease, and gastrointestinal disease. Sever-ity of disease measures include the number ofhospital days, blood transfusions, and vascularaccess procedures. Patients are followed untildeath, change of modality, transplant, loss-to-follow-up, or December 31, 1999, and are cen-sored at change of modality, transplant, loss-to-follow-up, or December 31, 1999.

Figures 8.18–27 present relative risks of death byhematocrit level or URR for 1998 incident hemo-dialysis patients. The data set construction,patient population, variable definitions, and ana-lytical methods are the same as those used in Fig-ures 5.21–30.

Table 8.2 presents the relative risks of death byrace for patient characteristics, comorbidity,disease severity, hematocrit, and URR, while Table8.3 presents equivalent data by diabetic status. Thedata set construction and methods are same asthose used in Tables 5.1 and 5.2.

Reference Section HPATIENT POPULATION

Counts of deaths (H.1), adjusted death rates(H.2–6), standardized mortality ratios (SMRs,H.7), and cause-specific death rates (H.8–13 andH.a.1–4) are reported for prevalent cohorts of1997, 1998, and 1999. Adjusted first-, second-, andthird-year death rates for incident cohorts are

reported in Tables H.14–16. Residents of the 50states, the District of Columbia, Puerto Rico, andthe Territories are included in each of these tables,as are all non-Medicare patients.

Tables H.1, H.8–13, H.a.1–4, and H.14–16include all causes of death. Tables H.2–6 excludepatients who died of AIDS. While patients whodied of street drug overdoses or accidentsunrelated to treatment are not counted in therates, their time at risk is counted until death.

Tables H.2–13 include both incident andprevalent patients. As defined earlier, prevalentcohorts include those patients who are alive onrenal replacement therapy at the beginning of theyear and whose first service date is at least 90 daysbefore the beginning of the year. Incident cohortsare limited to those patients who reach day 91 ofESRD treatment during the year. Because thesecalculations include only one-year of follow-up,a prevalent patient surviving until the end of theyear contributes one year at risk, while a prevalentpatient dying during the year contributes less thanone year. Since the calculation for incidentpatients begins on day 91 of ESRD, most of thesepatients contribute less than one year at risk; afull year is contributed only if day 91 of ESRD isJanuary 1 and the patient survives until the endof the year. Patients considered lost-to-follow-upat the beginning of the year are excluded fromthe analysis. The period at risk is not censored at

Table a.2Collapsed cause of deathcategoriesused for Reference Tables H.8–12

Collapsed categories Individual categoriesAcute myocardial infarction Myocardial infarction, acuteHyperkalemia HyperkalemiaPericarditis Pericarditis, including cardiac tamponadeAtherosclerotic heart disease Atherosclerotic heart diseaseCardiomyopathy CardiomyopathyCardiac arrhythmia Cardiac arrhythmiaCardiac arrest Cardiac arrest, cause unknownValvular heart disease Valvular heart diseasePulmonary edema Pulmonary edemaCerebrovascular disease Cerebrovascular accident including intracranial hemorrhage; ischemic brain damage/

anoxic encephalopathyG. I. hemorrhage Hemorrhage from transplant site; hemorrhage from vascular access; hemorrhage from dialysis

circuit; hemorrhage from ruptured vascular aneurysm; hemorrhage from surgery; otherSepticemia Septicemia, due to peritonitis; septicemia, due to peripheral vascular disease, gangrene;

septicemia, otherPulmonary infection Pulmonary infection (bacterial); pulmonary infection (fungal); pulmonary infection (other);

tuberculosisViral infection Viral infection, CMV; viral infection, other; Hepatitis B; other viral hepatitisAIDS AIDSOther infection Infection, other; fungal peritonitisCachexia CachexiaMalignant disease Malignant disease, patient ever on immunosuppressive therapy; malignant diseaseOther cause Pulmonary embolus; mesenteric infarction/ischemic bowel; liver-drug toxicity; cirrhosis;

polycystic liver disease; liver failure, cause unknown or other; pancreatitis; perforation of peptic ulcer; perforation of bowel; bone marrow depression; dementia, including dialysisdementia, Alzheimer’s; seizures; diabetic coma, hyperglycemia, hypoglycemia; chronicobstructive pulmonary disease (COPD); complications of surgery; air embolism; accident related to treatment; accident unrelated to treatment; suicide; drug overdose (street drugs);drug overdose; other identified cause of death

Unknown cause UnknownMissing forms Missing forms


the start of a lost-to-follow-up period, however;if a patient enters the lost-to-follow-up categoryduring a calendar year, he or she remains in thedeath rate computation until the end of that year.

Patient cohort populations often overlap. Patientswith a functioning transplant on the start day,for example, are included in the all-ESRD and allfunctioning transplant categories, while patientson dialysis are defined as both all-ESRD and all-dialysis. A patient in the all-dialysis category mayalso be reported in one of two subgroups—hemodialysis or CAPD/CCPD—if he or she hasbeen on that modality for at least the previous 60days. Dialysis patients who are not on hemo-dialysis or CAPD/CCPD, or who have been onthat modality for less than 60 days, are includedonly in the all-ESRD and all-dialysis categories.

Both adjusted and unadjusted death rates forprevalent cohorts are reported for the followinggroups (definitions are the same as those used inthe hospitalization analyses; see the discussion ofSection E):¨ all-dialysis; if a transplant occurs during

or at the end of the year the period at riskis censored at the transplant date

¨ hemodialysis; if a transplant occurs duringor at the end of the year the period at riskis censored at the transplant date

¨ CAPD/CCPD; if a transplant occursduring or at the end of the year the periodat risk is censored at the transplant date

¨ functioning transplant: patients with afunctioning transplant at the start of theperiod and who have had the transplantfor at least 60 days; the period at risk iscensored only at the end of the year

¨ all-ESRD; the period at risk is censoredonly at the end of the year

Patient populations for tables H.14–16 are thesame as those used in Reference Section I. Thepopulation groups include all dialysis, hemodi-alysis, CAPD/CCPD, and first transplant (knowncadaver and living only).

METHODS

Generalized mixed models are used to calculatethe smoothed rates in Tables H.2–6; these meth-ods are described later in this Appendix, as is themethod used to calculate the standardized mor-tality ratios (SMRs) in Table H.7. Death rates arereported for patient subgroups defined by age(nine groups), gender, race (white, black, NativeAmerican, and Asian), and primary diagnosis(diabetes, hypertension, glomerulonephritis, andother). Patients whose gender or date of birth is

missing, or who are of other races, are excludedfrom these patient populations, while patientswith no listed diagnosis are included in the “other”diagnosis group.

In Tables H.8–13 death rates are reported byprimary cause of death among patients prevalentat the beginning of, or incident (defined as 90days following the start of ESRD) during 1997,1998, and 1999. Subgroups are characterized byage, gender, race, and modality at the beginningof each cohort year for prevalent patients or at 90days of ESRD for incident patients. Dialysispatients are censored at transplant or the end ofthe calendar year, while transplant patients andpatients in the all-ESRD category are censoredonly at the end of the calendar year. The deathrate for a specific primary cause of death in eachsubgroup is obtained by dividing the total deathsfrom the primary cause by the subgroup’s totalfollow-up time. The sum of death rates for eachspecific cause in a subgroup is equal to the overalldeath rate of that subgroup. Death rates forcollapsed categories of death (Table a.1) arepresented in Tables H.8–12, while Tables H.a.1–4 list rates for each specific cause of death. TableH.13 presents rates by cause of withdrawal.

In Tables H.14–16 the adjusted first-, second-,and third-year death rates for incidentcohorts—including all-dialysis, hemodialysis,CAPD/CCPD, and first transplant patients—are computed from the Cox model, describedlater in this appendix. First-year rates arecomputed from the 1989–1998 populations foreach modality, second-year rates from the1988–1997 populations, and third-year ratesfrom the 1987–1996 populations. These deathrates are presented using aggregate categoriesfor age, gender, race, and primary disease, anda death rate presented for one of these variablesis adjusted for the remaining three. Overalldeath rates for all patients are adjusted for eachof the four variables; the death rates forHispanic and non-Hispanic groups, however,are unadjusted raw rates. The reference is thedistribution of age, gender, race, and primarydiagnosis for each cohort.

Reference Section IPATIENT POPULATION

These tables, which include only incident cohorts,present patient counts, counts of first renal trans-plants, and patient survival probabilities. Allcauses of death are included, as are all non-Med-icare patients and patients living in the 50 states,the District of Columbia, Puerto Rico, and theTerritories. Patients with unknown gender or age,


or whose age is listed as >110, are excluded fromthe patient cohorts.

Patient selection criteria are the same for bothunadjusted and adjusted survival probabilities. Allnew ESRD patients with a first ESRD service date(for dialysis or transplant) between January 1,1979, and December 31, 1998, are included in theanalysis. These patients are followed until Decem-ber 31, 1999, a maximum follow-up time of 15years and a minimum of one year.

Results are reported for the following groups:¨ all-ESRD: all ESRD patients beginning

renal replacement therapy in a calendaryear and surviving beyond day 90;patients are censored only at the end offollow-up

¨ 65 and over at start of ESRD: all ESRDpatients age 65 and over who begin renalreplacement therapy in a calendar year;patients are grouped in two-year periodsto increase cell size, and are censored onlyat the end of follow-up

¨ dialysis only: all ESRD patients startingrenal replacement therapy in a calendaryear, surviving beyond day 90, and notreceiving a transplant by day 91; patientsare censored at transplant or at the end offollow-up

¨ first renal transplant (cadaveric): patientsreceiving their first transplant in a calendaryear and for whom the donor is cadaveric

¨ first renal transplant (living): patientsreceiving their first transplant in a calendaryear and for whom the donor is living

In both transplant categories, patients for whomthe donor type is other or unknown are excluded.The cohort is defined by the year of first trans-plant, regardless of the year of first ESRD service.These patients are followed from the date of trans-plant (the date at which age is computed), andare censored only at the end of follow-up.

METHODS

Unadjusted patient survival probabilities areestimated using the Kaplan-Meier method, while

the Cox model is used for the adjustedprobabilities. Unadjusted survival probabilitiesfor Hispanics are new to this report.

To avoid excessive imprecision of the estimatedsurvival probabilities due to small cell sizes, theseprobabilities are presented using aggregatecategories for age, gender, race, and primarydisease, and a probability presented for one ofthese variables is adjusted for the remaining three.Overall probabilities for all patients are adjustedfor each of the four variables, as described laterin the discussion of statistical methods. Thereference is the distribution of age, gender, race,and primary diagnosis for each cohort, and thereferences are displayed in Table a.3.

CARDIOVASCULAR SPECIAL STUDIES ·CHAPTER NINEThis chapter—new to this year’s edition of theADR—addresses cardiovascular-relatedcomorbidities for ESRD patients, the incidenceof cardiovascular disease after the onset of ESRD,the use of evaluation tests before and after thefirst index of AMI, and the use ofrevascularization procedures following AMI indialysis patients. Analyses of cardiovascular-related comorbidities are based on all ESRDpatients, including non-Medicare patients, whileanalyses of the incidence of cardiovascular eventsand the use of evaluation procedures arerestricted to incident dialysis patients who areMedicare-eligible on day 91 of ESRD, limitingthe impact of non-Medicare patients withincomplete claims data. The start date of incidentdialysis patients, day 91 of ESRD, must fall afterthe start date based on Medicare paid dialysisclaims—the first day of the first month with atleast $675 of Medicare paid dialysis claims. If apatient’s start date does not fall on this date, heor she is excluded from the analyses. A patient isalso excluded if he or she did not survive at least90 days after the first ESRD service date,underwent transplantation within 90 days ofESRD, and/or had missing age, gender, or raceinformation. Missing age is defined as a missingdate of birth or an age calculated to be less thanzero or greater than 110.

Table a.3Reference population forSection I

One-year survival Two-year survival Five-year survival Ten-year survivalAll ESRD 1988-98 all ESRD 1987-97 all ESRD 1984-94 all ESRD 1979-89 all ESRD65 and over 1989-97 65+ ESRD 1987-97 65+ ESRD 1983-93 65+ ESRD 1979-89 65+ ESRDDialysis only 1988-98 dialysis 1987-97 dialysis 1984-94 dialysis 1979-89 dialysisFirst transplant 1988-98 first transplant 1987-97 first transplant 1984-94 first transplant 1979-89 first transplant (cadaver) (cadaver) (cadaver) (cadaver) (cadaver)First transplant 1988-98 first transplant 1987-97 first transplant 1984-94 first transplant 1979-89 first transplant (living) (living) (living) (living) (living)


Figures 9.1–4 present, for different populations,the distribution of cardiovascular-relatedcomorbidities, including atherosclerotic heartdisease (ASHD), congestive heart failure (CHF),cerebrovascular accident/transient ischemicattack (CVA/TIA), peripheral vascular disease(PVD), and other cardiac disease. The firstoccurrence date of each comorbidity is obtainedfrom Part A institutional claims, REBUS inpatientdata, and Part B physician/supplier claims.Comorbid conditions at initiation are determinedfrom the Medical Evidence Form (HCFA 2728),in which CHF, CVA/TIA, and PVD can be founddirectly, ischemic heart disease, myocardialinfarction, and cardiac arrest are grouped intoASHD, and cardiac dysrhythmia and pericarditisare combined into other cardiac disease.

Figures 9.1 and 9.2 compare patient comorbidityreported on the Medical Evidence Form at theinitiation of ESRD service and that obtained fromthe two-year pre-ESRD claims data. A total of82,508 incident ESRD patients from 1995 to 1997were identified as being 67 years old or older atinitiation, having claims information within twoyears prior to ESRD, and showing validcomorbidity information on the MedicalEvidence Form. For each of the incident cohorts,the percent of patients who had ASHD, CHF,CVA/TIA, PVD, and/or other cardiac disease arecomputed from both the Medical Evidence Formand the claims data.

Figures 9.3 and 9.4 present the distribution of co-morbidity in incident and prevalent ESRDpatients, respectively. The incident populationincludes ESRD patients from 1995 to 1998 whosurvived at least nine months after the onset ofESRD, while the point prevalent populationincludes ESRD patients from the same timeperiod who began ESRD service at least ninemonths before the point prevalent date. The fig-ures show the percent of patients with claims forcardiovascular disease as of the end of the ninthmonth after ESRD (incident patients) or as of thepoint prevalent date.

Figures 9.5–22 describe the occurrence of cardio-vascular disease among incident Medicare dialy-sis patients after the onset of ESRD. Patients areclassified as having a particular disease as of thefirst date a related code appears in the claims: forarrhythmia, valvular disease, and cardiomy-opathy, the date the ICD-9-CM diagnosis codefirst appears in Part A institutional and/or Part Bphysician/supplier claims files; for newly diag-nosed CVA and TIA, the date the code firstappears in Part A inpatient claims data to iden-

tify either a principal or secondary diagnosis. AMIand cardiac arrest are identified through ICD-9-CM diagnosis codes for the event or a death dueto the event (codes for AMI appear in Part A in-patient claims, while those for cardiac arrest ap-pear in Part A and/or Part B claims). Coronaryrevascularization is defined through ICD-9-CMprocedure codes in Part A institutional claimsfiles, while peripheral revascularization is definedthrough ICD-9-CM procedure codes in Part Aclaims and/or Current Procedural Terminology(CPT) codes in Part B claims data. Major ampu-tation is identified through CPT codes in Part Bphysician/suppliers claims data. The codes usedto identify patients with cardiovascular diseaseare as follows:

¨ arryhthmia: 426–427¨ valvular disease: 394–396, 397.0, 424.0–

424.3¨ cardiomyopathy: 425¨ cardiac arrest: 427.5¨ AMI: 410, 410.X0, 410.X1¨ newly diagnosed CVA: 430–434, 436–437¨ TIA: 435¨ coronary revascularization: 36.01, 36.02,

36.05, 36.06, 36.1X¨ peripheral revascularization: 39.25, 39.29,

35452, 35454, 35456, 35458, 35459, 35470,35473–35475, 35482–35485, 35492–35495, 35511, 35516, 35518, 35533, 35541,35546, 35548, 35549, 35551, 35556, 35558,35563, 35565, 35566, 35571, 35582, 35583,35585, 35587, 35612, 35616, 35621, 35623,35641, 35646, 35651, 35654, 35656, 35661,35663, 35665, 35666, 35671

¨ major amputation: 23900, 23920, 24920,24900, 25900, 25905, 25920, 25927, 27295,27590, 27591, 27592, 27598, 27880, 27881,27882, 27888, 27889, 28800, 28805

Figures 9.5–10 address cardiac disease amongpediatric ESRD patients. Data for these figuresinclude children under 20 years old who wereincident Medicare dialysis patients from 1991 to1996. Patients are followed from day 91 of ESRDuntil the first of death, transplant, loss-to-follow-up, or the end of 1999. Figures 9.8–10 present, byage, gender, and race, the incidence of cardiacarrest, valvular disease, arrhythmia, andcardiomyopathy during the first three-yearfollow-up period. Figure 9.7 shows the adjustedfirst three-year annual rate per 1,000 patient-yearsat risk for cardiac arrest, cardiac death,arrhythmia, and cardiomyopathy. The methodof calculating adjusted rates is described in thediscussion of statistical methods later in thisappendix. The stratum in the Cox regressionmodel is incident year, and covariates include age,


gender, race, and primary cause of renal failure.The reference is the population distribution ofall included patients from 1991–1996.

Figures 9.11–22 present demographic data andadjusted rates per 1,000 patient-years at risk forall-cause mortality, cardiac morbidity andmortality, and combined events among incidentMedicare dialysis patients from 1991 to 1998.Patients are followed from day 91 of ESRD to theearliest of death, transplant, loss-to-follow-up, orthe end of 1999. The method used to computethe adjusted cardiovascular-related event rate isdescribed later in the discussion of statisticalmethods. The stratum in the Cox regressionmodel is incident year, and covariates include age,gender, race, and primary cause of renal failure.The reference is the population distribution ofall included patients from 1991–1998 for the first-year event rate; from 1991–1997 for the second-year event rate; and from 1991–1996 for the third-year event rate.

Figures 9.23–28 describe the use of evaluationprocedures in patients who had an AMI after theonset of ESRD. “Any stress test” is defined as thefirst stress echo, stress nuclear test, or stress ECG(based on the CPT codes in Part B physician/sup-plier claims), or a stress test (identified throughICD-9-CM procedure codes in Part A institu-tional claims data). Coronary angiography and/or catheterization are identified through the ICD-9-CM procedure codes in Part A institutionalclaims or CPT codes in Part B claims files. Rest-ing echocardiography tests are identified throughCPT codes in Part B physician/supplier claimsdata, and lipid measurement through CPT codesin Part A institutional revenue claims or Part Bphysician/supplier claims data. The codes usedto identify patients with evaluation tests pre- andpost-AMI are as follows:

¨ any stress test: 89.41–89.44, 93350, 78459,78460, 78461, 78464, 78465, 78472, 78473,78478, 78480, 78481, 78483, 78491,78492,93015–93018

¨ coronary angiography/catheterization:88.53–88.57, 37.21–37.23, 93508, 93510,93511, 93524, 93526, 93527, 93529, 93531,93532, 93533, 93539, 93540, 93543, 93545,93555, 93556

¨ resting echocardiography: 93303, 93304,93307, 93308, 93312, 93320, 93321, 93325

¨ lipid measurement: 80061, 82465, 84478,83715–83721

Figures 9.23(a)–28(a) present data on evaluationprocedures and survival status within 30 days (sixmonths for lipid tests) of AMI. All incident Med-

icare dialysis patients with AMI are divided intogroups as follows, and the percent distribution isshown in the figures:

¨ patients receiving the procedure within 30days and surviving longer than 30 days

¨ patients receiving the procedure beforethey died, underwent transplant, or werelost-to-follow-up within 30 days follow-ing AMI

¨ patients not receiving the procedure within30 days and surviving longer than 30 days

¨ patients not receiving the procedure be-fore they died, underwent transplant, orwere lost to follow-up within 30 days fol-lowing AMI

For patients who survived at least 30 days afterAMI, the comparisons of patients receiving theevaluation procedure in the month before andthe month following AMI are presented in Fig-ures 9.23(b)–26(b). The pre-AMI evaluation pro-cedure is the latest one before the onset of AMI,while the post-AMI procedure is the first one fol-lowing that onset. Among the total number ofcoronary revascularization procedures usedwithin 30 days after AMI, the percent use of per-cutaneous transluminal coronary angioplasty,stent, and coronary artery bypass are computedand presented in Figure 9.28(b).

PREVENTIVE HEALTH CARE MEASURES ·CHAPTER TENMethods for determining screening rates forbreast and cervical cancer, diabetic eye exams, andglycosylated hemoglobin testing (HbA1c) aretaken directly from HEDISÒ 2000 specifications(HEDISÒ 2000 is a program of the NationalCommittee for Quality Assurance, and is used tomonitor the performance of managed health careplans). Because HEDISÒ 2000 does not addressprostate cancer or influenza vaccinations,algorithms for these analyses have been createdby the USRDS. Screening rates are determinedfor both incident and prevalent ESRD patients.

Screening intervals for cervical and prostatecancer include the reporting year and the twoprior years; for breast cancer, the reporting yearand the prior year; for glycosylated hemoglobin,the reporting year only; for diabetic eye exams,the reporting year and the prior year if diabeteswas not the cause of renal failure, otherwise, thereporting year only; and for influenzavaccinations, September 1 through December 31of the reporting year.

Patients with Medicare as a secondary payor ornot eligible for Medicare are omitted from all


analyses. Also omitted are those who have a miss-ing date of birth, who did not survive the entirereporting year, who were ESRD less than 90 daysprior to January 1 of the reporting interval, and,for diabetic eye exams and glycosylated hemo-globin screening, who are non-diabetic.

For the analysis of prostate cancer screening, pa-tients are excluded if their claims contain any ofthe following ICD-9-CM procedure codes: 60.2,60.21, 60.29, 60.3, 60.4, 60.5, or 60.62. Codes usedto identify patients who receive screening includeCPT-4 codes of 52601, 52612, 52614, 55801,55810, 55812, 55815, 55821, 55831, 55840, 55842,55845, and 84153; revenue codes of 0300 or 0310,associated with an ICD-9-CM diagnosis code of185 or 233.4; and ICD-9-CM procedure codesof 60.11, 60.12, 60.18, 87.92, and 91.39.

In Figures 10.1 and 10.9 the numerator includesall patients receiving an influenza vaccination inthe last four months of 1999, while the denomi-nator includes all prevalent patients. Patients whoreceive flu vaccinations are identified throughCPT codes of 90724, 90657, 90658, and 90659,and a HCPCS code of G0008.

Figures 10.2 and 10.6–10.8 display rates of devel-opment of cancers per 1,000 ESRD patient-years.ICD-9-CM codes used for breast cancer are:198.81, 233.0, 174.XX, and 175.XX; for cervicalcancer: 198.82, 233.1, and 180.XX; for prostatecancer: 198.82, 233.4, 233.9, and 185.XX.

Figure 10.12 shows the likelihood of hospitaliza-tion in January through March 1999 by whetheror not a patient received an influenza vaccina-tion from September through December 1998.Calculated percentages are adjusted for age, gen-der, race, prior ESRD time, and total hospital daysduring January through August 1998.

Figures 10.17–18 describe the use of diabetictesting supplies during 1999. HCPCS codes usedto identify these supplies are A4250, A4253–55,E0607, E0609.

Figures 10.11 and 10.19–22 display lipidmonitoring in diabetic and non-diabetic patients.CPT codes used to identify lipid monitoring are80061, 82465, 83715–721, and 84478.

PROVIDER CHARACTERISTICS ·CHAPTER ELEVEN & REFERENCE SECTION JThese sections contain data from HCFA’s AnnualFacility Survey, the CDC National Surveillanceof Dialysis-Associated Diseases, Freedom ofInformation Act requests regarding unit chain

affiliation, and Public Use Files of cost reports forrenal dialysis units (provided on HCFA’s website).Data from facilities that have returned a HCFAsummary and/or a CDC survey are summarizedin the tables and figures. Because the CDC didnot conduct a survey in 1998, data in someinstances is reported only up to 1997.

For these analyses, a chain-affiliated unit is de-fined as one of a group of 20 or more freestand-ing dialysis units owned by a common party andlocated in more than one state.

ECONOMIC COSTS OF ESRD ·CHAPTER TWELVE & REFERENCE SECTION KChapter TwelveData in this chapter are calculated using severaldifferent analytical models. New to this year’sADR is the use, in the majority of the economicanalyses, of the as-treated model, which replacesthe intent-to-treat model used in previous edi-tions. The as-treated model is described in detailin the discussion of Reference Section K, below.

HCFA MODEL

This model, described in the HCFA researchreport on ESRD (1993–1995), is used for Tables12.6–8. With this method patients are classifiedinto four mutually exclusive treatment groups:

¨ dialysis: ESRD patients who are on dialy-sis for the entire calendar year, or for thatpart of the year in which they are alive,ESRD, and Medicare entitled

¨ transplant: ESRD patients who have akidney transplant during the calendar year

¨ functioning graft: ESRD patients whohave a functioning kidney transplant forthe entire calendar year, or for that part ofthe year in which they are alive, ESRD, andMedicare entitled

¨ graft failure: ESRD patients who have hada transplant, but returned to dialysis dueto loss of graft function during the calen-dar year; patients with a graft failure and atransplant in the same calendar year arealways classified in the transplant category

Patients are categorized as having Medicare assecondary payor on the basis of the “Primarypayor amount” on Part A and Part B claims.

SIX-MONTH ENTRY MODEL

This model considers prevalent patients whosurvive on a given modality during the entryperiod (for these analyses, the first six monthsof 1998 and 1999). Patients are characterizedusing Medicare claims data from the entryperiod, and costs are aggregated for the follow-


up period (the last six months of 1998 and1999), with patients censored at modalitychange, death, loss-to-follow-up, or the end ofthe follow-up period (December 31, 1998 or1999). Patients who were MSP during 1998 or1999 are excluded.

METHODS

Table p.1 in the Précis summarizes data on thecosts of ESRD treatment. Total 1999 Medicarespending is calculated from the claims data,and includes all paid claims for ESRD patientsin the USRDS database. Cost aggregation be-gins at the first ESRD service date for eachpatient. Total 1999 Medicare spending is in-flated by 2% to account for incomplete claims,and organ acquisition costs are estimated withthe same methods used in the 1999 ADR (149–150). HMO costs are estimated using the totalHMO months for 1999, obtained from theHCFA managed care organization file, in con-junction with the 1999 AAPCC rate.

Non-Medicare spending by Employee GroupHealth Plans (EGHPs) is estimated by computingthe per year at risk costs for EGHP and non-EGHP patients separately, then multiplying thedifference by the EGHP years at risk for 1999.Patient obligations are estimated as 18% of thesum of Medicare payments, non-federal EGHPcosts, and patient obligations (1999 ADR, 149).Because non-Medicare patients are estimated toconstitute 7% of all ESRD patients in the U.S.(1999 ADR, Table ES-1), costs are estimated as7% of the total costs of Medicare patients.

Changes in Medicare spending from 1998 to 1999are taken directly from Table K.1, without the 2%adjustment for late claims. Per patient year at riskfigures are based on patients who were never MSPduring the study period (Tables K.18–19), againusing non-inflated results. The apparent decreasein per patient per year costs is likely artifactual,and due to the absence of late claims affectingthe 1999 claims dataset. The range for inflation-adjusted costs is calculated using the overallConsumer Price Index (2.7%) as well as theMedical Consumer Price Index (3.7%). The year-at-risk figures for modality during the 1995–1999time period are taken directly from Table K.4;these figures include non-MSP patients only, andare not adjusted for late claims.

PMPM payments by hematocrit and URR arecalculated using the six-month entry model.Data are obtained from outpatient dialysisclaims, and only those patients with four ormore EPO claims (Figures 12.10–11 and

12.13–14) or three or more claims with validURR data (Figures 12.12–12.14) are included.The hematocrit range for each patient is basedon the mean hematocrit for the six-monthentry period, while the URR range is based onthe median value from the same period. Claimsare aggregated by month, and in the case ofmultiple claims per month only the last valueof the month is used. The allowable costspresented in Figures 12.11–12 are adjustedusing an ordinary least squares regression oflog transformed PMPM with smearingestimate.

Information about the construction of otherfigures and tables is provided in the captions.

Reference Section KMEDICARE CLAIMS DATA

The cost information in this section is derivedfrom Medicare Part A and Part B (physician/supplier) claims data on the HCFA StandardAnalytic Files, which are created annually sixmonths after the end of each calendar year. Thedata for 1995 to 1999 comprise approximately28 million institutional claims (hospitalinpatient and outpatient facilities, outpatientdialysis facilities, skilled nursing facilities,hospice facilities, and home health agencies),as well as 213 million line items from physician/supplier claims. Claims data are obtained forall patient ID numbers in the USRDS database,and the Renal Beneficiary Utilization System(REBUS) is used to gather all HCFA IDnumbers under which patients may haveclaims. The claims data are then merged withpatient demographic data and modalityinformation in the USRDS database.

The economic analysis for this reference sectionfocuses on two amounts found in the claims data:the claim payment amount, which is the amountof payment made from the Medicare trust fundfor the services covered by the claim record; andthe pass-through per diem amount, which appliesto inpatient claims and reimburses the providerfor capital-related costs, direct medical educationcosts, and kidney acquisition costs. ChapterTwelve includes another dollar amount calledMedicare Allowable, defined here as the amountof allowed charges for the services covered by theclaim record. For institutional claims, this amountis calculated as the sum of the amounts from theMedicare payment, coinsurance, deductible, andany payment provided by a payor other thanMedicare. For physician/supplier claims, theMedicare Allowable amount is provided byHCFA as a separate data element.


PAYMENT CATEGORIES

Medicare payments are broken down into sev-eral categories, as shown in Table a.4. New thisyear, estimates of costs from the Outpatient SAFare derived for the individual services provided.Since actual payment amounts are provided onlyfor the entire claim, cost estimates for dialysis,EPO, iron and so forth are calculated from theclaim-level “Total Charge,” the payment amount,and the revenue line-level “Total Charge,” as fol-lows: payment (line) = [total charge (line) / totalcharge (claim)] * payment (claim).

AS-TREATED MODEL

In an as-treated model patients are initiallyclassified by their modality at entry into theanalysis, and they retain that classification untila change in modality occurs. When a modality

change is encountered, the beginning modalityis censored at the change date plus 60 days, anda new observation with the new modality iscreated. The first 60 days after a modalitychange are attributed to the previous modalityin order to account for any carryover effectsrelated to that modality. If the modality changeis from dialysis to transplantation, the 60-daycarryover period is not added to the dialysismodality, which is censored (and the transplantmodality begins) on the date of the transplanthospital admission. In the case of changesinvolving only a new dialysis modality, the newmodality must last at least 60 days in order tobe counted. Aggregation of Medicare paymentsis done on an as-treated basis, attributing allpayments to the patient’s modality at the timeof the claim.

Table a.4Medicare categories ofpayment

Medicare payment categories Basis for categorizing claimTotal Sum of all payments

Total inpatient Sum of all payments originating from the inpatient SAF, including pass-throughsMedical DRG Inpatient SAF, DRGSurgical DRG Inpatient SAF, DRGTransplant DRG Inpatient SAF, DRG 302Other DRG Inpatient SAF, DRG not included in the above categoriesNon-transplant pass-throughs Inpatient SAF, DRG not 302, calculated from per diem and covered daysTransplant pass-throughs Inpatient SAF, DRG 302, calculated from per diem and covered days

Total outpatient Sum of all payments originating from the Outpatient SAFOutpatient hemodialysis Outpatient SAF, hemodialysis revenue codesOutpatient peritoneal dialysis Outpatient SAF, peritoneal dialysis revenue codesOutpatient other dialysis Outpatient SAF, dialysis revenue codes other than HD or PDOutpatient EPO Outpatient SAF, revenue codes and/or HCPCS codeOutpatient Calcijex Outpatient SAF, revenue and HCPCS codesOutpatient iron Outpatient SAF, revenue and HCPCS codes

Outpatient other injectables Outpatient SAF, revenue and HCPCS codesRadiology Outpatient SAF, revenue and/or CPT codesPharmacy Outpatient SAF, revenue codesAmbulance Outpatient SAF, revenue codesLaboratory/pathology Outpatient SAF, revenue and/or CPT codesOutpatient other Outpatient SAF, does not qualify for any other cost category

Skilled nursing facility Skilled nursing facility SAFHome health agency Home health SAFHospice Hospice SAF

Total physician/supplier Sum of physician/supplier paymentsTransplant surgery Physician/supplier SAF, CPT codesInpatient surgery Physician/supplier SAF, CPT and place of service codesOutpatient surgery Physician/supplier SAF, CPT and place of service codesE&M nephrologist inpatient Physician/supplier SAF, CPT, place of service and specialty codesE&M nephrologist outpatient Physician/supplier SAF, CPT, place of service and specialty codesE&M non-nephrologist inpatient Physician/supplier SAF, CPT, place of service and specialty codesE&M non-nephrologist outpatient Physician/supplier SAF, CPT, place of service and specialty codesDialysis capitation Physician/supplier SAF, CPT and/or type of service codesInpatient dialysis Physician/supplier SAF, CPT codesHome dialysis Physician/supplier SAF, HCPCS and place of service codesVascular access Physician/supplier SAF, CPT codesPeritoneal access Physician/supplier SAF, CPT codesPhysician/supplier EPO Physician/supplier SAF, HCPCS codesPhysician/supplier iron Physician/supplier SAF, HCPCS codesImmunosuppressive drugs Physician/supplier SAF, HCPCS codesDurable medical equipment Physician/supplier SAF, HCPCS codesPhysician/supplier radiology Physician/supplier SAF, CPT and specialty codesPhysician/supplier lab/path. Physician/supplier SAF, CPT codesPhysician/supplier ambulance Physician/supplier SAF, HCPCS and place of service codesOther physician/supplier Physician/supplier SAF, does not qualify for any other category


Patients are classified in these tables into four as-treated modality categories: hemodialysis,CAPD/CCPD, other dialysis, and transplant. The“other dialysis” category includes cases in whichthe dialysis modality is unknown or is nothemodialysis or CAPD/CCPD, while the“transplant” category includes patients who havea functioning graft at the start of the period orwho receive a transplant during the period. Sometables also include categories for all-dialysis(hemodialysis, CAPD/CCPD, and other dialysis)and all-ESRD (all-dialysis and transplant).

The study spans the five years from January 1,1995 to December 31, 1999, and ESRD patientsprevalent on January 1, 1995 or incident at anytime during the period are potentially eligible forinclusion. The initial study start date for a givenpatient is defined as the latest of the following:

¨ January 1, 1995¨ thirty days after the first ESRD service date

in the USRDS database for that patient¨ for dialysis patients, 30 days after the first

month in which dialysis payments exceed$675

Because it is impossible to characterize the totalcost of their care, patients for whom Medicare isthe secondary payor (MSP) at any time during thestudy period, as determined from the Medicare En-rollment Database, are excluded from the analysis.

Medicare payments are aggregated from themodality start date until the earliest of death,transplant, modality change, loss-to-follow-up,or December 31, 1999 for each modality period.Dialysis patients are defined as lost-to-follow-upafter a period of three consecutive months inwhich dialysis payments (institutional plusphysician/supplier) fall below $675/month, andpatients incurring no Part A or Part B Medicarecosts for the entire period are excluded. Medicarepayment amounts are linearly prorated for claimsthat span the start or end date of a modalityperiod or of the study itself.

In order to express the costs as dollars per yearat risk (YAR), total costs during the follow-upperiod are divided by the length of the follow-up period. Costs per year at risk are calculatedby patient category, and stratified by age,gender, race, modality, and diabetic status.Diabetic status is based on the primary diseasecausing ESRD, as recorded on the MedicalEvidence form. A patient with a non-diabeticcause of renal failure may have diabetes, butthe disease is not judged to be the cause ofESRD. Patient age is calculated at the study start

date, and patients with a missing date of birthare excluded from the analysis.

INTERNATIONAL COMPARISONS ·CHAPTER THIRTEENThe international dialysis and transplant data for1999 have been collected from the followingcountries using a data form designed by theUSRDS: the Australian and New Zealand Dialysisand Transplant Registry (ANZDATA), the AustriaOEDTR, Belgium Etterbeek-Ixelles Etterbeek-Elsene, Bulgaria First Hemodialysis Centre, theCatalan Renal Registry, the Chilean RenalRegistry, the Czech Society of Nephrology, theEstonian Society of Nephrology, the FinnishRegistry for Kidney Diseases, the QuaSi-Niere inGermany, the Greek Hellenic Society ofNephrology, the Hungarian Transplant Registry,Landspitalinn (Iceland), the Israeli Renal Registry,the Japanese Society of Dialysis Therapy, theLatvian Renal Registry, the Norwegian RenalRegistry, the Society of Dialysis in Russia, theSingapore Renal Registry, the Swedish RenalRegistry, the United Kingdom TransplantSupport Service Authority, the Uruguay Dialysisand Transplant Renal Registry, the Institute ofNephrology in Yugoslavia, and the USRDS. Forall countries but the U. S., data prior to 1998 aretaken from the 1993–1998 Annual Data Reports;1998 data were collected for the 2000 ADR.

CENSUS POPULATION BASE ·REFERENCE SECTION LCensus data in this section, which are used tocalculate rates throughout the chapters andReference Tables, are obtained from the UnitedStates Census Bureau. Updated populationestimates are available at www.census.gov.

STATISTICAL METHODSMethods for adjusting ratesDIRECT ADJUSTMENT

There are several rate adjustment methods, butonly the direct method allows rates to be com-pared (Pickle LW, White AA). With this methodthe adjusted rate is derived by applying theobserved category-specific rates to one standardpopulation, i.e. the adjusted rate is a weightedaverage of the observed category-specific rates,using as the weight the proportion of each cat-egory in the reference population. This methodis used to produce some of the incident andprevalent rates in Reference Sections A and B.

MODEL-BASED ADJUSTMENT

There are, however, some disadvantages to thedirect adjustment method. If one category in agroup has only a few patients, for example, the


death rate for this category will not be stable,making the adjusted death rate unstable as well.In addition, if one category in a group containsno patients, the adjusted death rate for this groupcannot be calculated. A model-based adjustmentis needed to overcome these disadvantages. Inprevious years we have simply run the modelsand substituted the average values from thereference for the covariates. Unless the model islinear, however, averaging on the covariates is notthe same as averaging on the model. In this ADRwe thus use models to predict death rates for eachcategory in each group, and then use the directadjustment method to calculate adjusted deathrates based on predicted death rates with a givenreference population. Standard errors arecalculated using the bootstrap approach.

This method is used to calculate adjusted deathrates, event rates based on the Cox regressionmodel, and adjusted incident and prevalent ratesbased on the Bayesian spatial model. Adjustedsurvival probabilities are calculated by taking theexponential of negative adjusted death rates.

Because different adjusted event rates for differentgroups have different reference populations, thesepopulations are identified in the relevant sections.

KAPLAN-MEIER UNADJUSTED SURVIVAL

PROBABILITIES

The Kaplan-Meier method and Greenwood’sformula (Kalbfleisch JD, Prentice RL) are used toestimate unadjusted survival probabilities andstandard errors in Chapters Six, Seven, and Eight,and Reference Sections G and I. Survivalprobabilities are expressed as percentages varyingfrom 0 to 100.

COX REGRESSION: ADJUSTED SURVIVAL

PROBABILITIES

Because of the different mix of patients each year,unadjusted survival probabilities may not becomparable across cohort years. Adjustedanalyses, however, make results comparable byreporting probabilities that would have arisen hadeach incident cohort contained the samedistribution of age, gender, race, and primarydiagnosis as the reference population. Adjustedsurvival probabilities, calculated using the model-based adjustment method described above, arereported in Reference Sections G and I, with age,gender, race, and primary diagnosis used asadjusting factors. Probabilities are estimated usingthe Cox regression model, stratified by year(Kalbfleisch JD, Prentice RL). Data are reportedfor all new ESRD patients, new ESRD patientsover age 65, and new dialysis, cadaveric transplant,

and living donor transplant patients. Graftsurvival rates are also reported for cadaveric andliving donor transplants. This process yieldsestimates of the survival probabilities that wouldhave arisen in each year for patients in thereference population. Since the probabilities ineach table are adjusted to the same type ofpatients, any remaining differences among yearsare due to factors other than age, gender, race,and primary diagnosis.

ADJUSTED INTERVAL DEATH RATES

Adjusted one-, two-, and three-to-five yearinterval death rates, reported for incident cohortsin Chapter Eight and in Tables H.14–16, areestimated using the Cox regression and themodel-based adjustment method. The intervaldeath rate is estimated as the difference ofcumulative death rates at the start and end of theinterval, while the annual death rate is estimatedby dividing the interval death rate by the lengthof the interval (in years), under the assumptionthat the death rates are constant over this period.Similar to the adjusted survival probabilities, theadjusted interval death rates are comparableacross years within this ADR. This method wasalso used to calculate adjusted death rates andother event rates in Chapter Nine.

Generalized mixed modelThe generalized mixed model with log link andPoisson distribution is used to calculate deathrates, first admission (hospitalization) rates, andfirst transplant rates. While rates are reported onlyfor 1999, three years of prevalent data (1997–1999) are used to improve the stability of the esti-mates.

The generalized mixed model, which considersboth fixed and random effects, is implementedusing the SASÒ macro GLIMMIX (SASÒ Version8). The Poisson rates for the intersections of age,gender, race, and diagnosis are estimated usingthe log linear equation Log (rate) = (fixed effects)+ (random effect). Fixed effect variables includeyear, age, gender, race, and diagnosis, and all two-way interactions between age, gender, race, anddiagnosis. Assumed to be independently andidentically distributed with a normal distribution,the random effect is the four-way interaction ofage, gender, race, and diagnosis.

For the 2001 ADR we used a single model tocalculate all rates in a single table for bothintersecting and marginal groups, rather than the16 models used previously. The marginal ratesare simply the weighted averages of the estimated,cross-classified rates, with cell-specific patient


years as weights. Because the use of a single modelmeans that GLIMMIX cannot give the standarderrors for these computed rates, and becauseGLIMMIX does not take all variations intoaccount for the intersecting group rates, we alsoused the bootstrap technique to calculate thestandard errors. This method was used tocalculate the prevalent death rates and firsthospital admission rates, along with theirstandard errors, in Tables E.1–5, F.19, and H.2–6.

Standardized mortality ratioThe standardized mortality ratio (SMR)measures the mortality rate for a group of patientsrelative to the expected death rate for the groupbased on the death rate of the referencepopulation.

In Table H.7, for example, SMRs are used tocompare death rates for prevalent dialysis patientsin each state from 1997 to 1999 to the nationaldeath rates of U.S. dialysis patients in 1999 (TableH.2). The SMR accounts for the age, gender, race,and diagnosis of the prevalent dialysis patients ineach state. The expected number of deaths in eachage, gender, and race intersection group of theobserved population is calculated by multiplyingthe group-specific standard rates by the totalfollow-up time at risk of the observed patients inthe group. The total expected number is thencalculated by summing the expected numbers ineach group, and the SMR is the ratio of theobserved to the expected number of deaths. AnSMR of 1.05 for a subgroup indicates that thisgroup has a risk of death approximately fivepercent higher than that of the referencepopulation.

Standardized first transplantation ratios (STRs)and first admission standardized hospitalizationratios (SHRs), calculated using similar methods,are reported in Tables F.24 and E.6.

Mapping methodsMapping is an important tool for assessingenvironmental determinants and illustratingspatial patterns and temporal trends. Geographicresolution is enhanced by mapping at the level ofsmall regions, but because this can increase datainstability, smoothing methods are needed tostabilize the data. The methods described herehave been used in the majority of maps presentedin the Annual Data Report. Because thedistribution of age, gender, and race in apopulation can affect ESRD incident andprevalent rates, we have also included maps thatare adjusted for these variables as well assmoothed.

The majority of disease mapping within the ADRis by Health Service Area (HSA), an approach wehave borrowed from the Atlas of United StatesMortality (Centers for Disease Control and Pre-vention). Each HSA is a group of counties de-scribed by the CDC authors as “an area that isrelatively self-contained with respect to hospitalcare.” In all maps HSAs have been divided equallyamong five ranges to help readers more easilyinterpret the information.

Throughout the ADR, data in maps and graphsare unadjusted unless noted. HSA-levelinformation is mapped according to the patient’sresidence, and, because of area size and limitationsin the mapping software, data for the District ofColumbia, Puerto Rico, and the U.S. Territoriesare not included in the maps.

METHODS FOR SMOOTHING & ADJUSTING DATA

To smooth data we use both the Bayesian spatialmodel (Waller LA, Carlin BP, Xia H, Gelfand AE)and the weighted head-banging method (TukeyPA, Tukey JW; Mungiole M, Pickle LW, SimonsonKH).

Bayesian spatial modelThe Bayesian method, used for all maps in theADR unless stated otherwise in the caption, is astatistical approach that uses the Poisson regres-sion model to fit, for example, the incident ratesof the regions. The relative risks for the regions,as random effects, follow the ConditionalAutoregression (CAR) Normal distribution, andthe precision of the relative risks has a Beta dis-tribution (Waller LA, Carlin BP, Xia H, GelfandAE). This model smoothes the incident rate byborrowing information for each HSA from itsneighbors through the relationship defined byCAR; neighbors, in our definition, are HSAs thatshare a boundary. The exponential offsets in themodel are the internally standardized incidentcounts. Smoothed incident rates are obtained bydividing the predicted counts by the correspond-ing population sizes. For smoothed and adjustedmaps, fixed effects of age, gender, and race wereused in the Bayesian model. The smoothed andadjusted incidents rates are also calculated usingthe model-based adjustment method for eachHSA, and are based on the smoothed rates foreach category in each HSA from the BayesianSpatial Hierarchical model, using the nationalpopulation as reference.

Weighted head-banging methodThis method for smoothing data in maps usesweighted medians from geographical neighbor-hoods. The surrounding neighbors of a given


region with value y (to be smoothed) are collectedand paired so that they are located geographicallyin as straight a line as possible, with the region ofinterest located in the center. The weighted me-dian l of the smaller values in all paired neigh-bors, and the weighted median u of the largervalues, are defined (Mungiole M, Pickle LW,Simonson KH). If y is between l and u a newsmoothed value for this region is assigned withvalue y; otherwise, it is assigned l or u or y,depending on their values and weights. This pro-cess is repeated iteratively until the differencesbetween the new smoothed value and its previ-ous value for all regions are within the pre-statedrange. The population size of the correspondingregion is usually used as the weight.

This method was used to smooth maps in Chap-ters Four, Eleven, and Twelve (its use is indicatedin the captions), along with other maps that couldnot fit the Poisson model.

MISCELLANEOUSSpecial studies & data collection formsCopies of the HCFA Medical Evidence form(2728) and Death Notification Form (2746);of the UNOS Transplant Candidate Registra-tion Form, Kidney Transplant Recipient Reg-istration Form, and Kidney Transplant Recipi-ent Follow-up Form; and of forms used fordata collection in past USRDS special studies,are available on the USRDS web site atwww.usrds.org.

CaptionsFigure captions in the ADR provide descriptionsof patient cohorts and data adjustment, alongwith other general information regarding thefigures, and should be read in conjunction withthe explanations provided in this appendix.

FUTURE DIRECTIONS OF THE USRDSTo improve and advance research on end-stagerenal disease, the USRDS and its biostatisticalteam plan to investigate the following issues.

Spatial smoothingHigh-resolution maps require smoothing in or-der to balance spatial focus with statistical stabil-ity. Because hierarchical models are more flex-ible and effective, the head-banging approach willbe eliminated in future analyses, and the Baye-sian hierarchical models adapted more fully tothe needs of the USRDS. In this year’s ADRcovariates are used as fixed effects for adjusting,and neighbors are defined as HSAs with com-mon borders. In the future, we will investigateother spatial correlation structures, for example,

including those which incorporate spatial dis-tance in the correlation structure.

State-level adjusted incident ratesAs mentioned in the section describing statisti-cal methods, if one subgroup of a populationcontains only a few people the adjusted rate maynot be stable. This instability occurred in our cal-culation of state-level adjusted incident rates(adjusted for age, gender, and race) due to thefine age groups and uneven distribution of racesacross the states. We have found that this prob-lem persists even with the use of combined agegroups and the removal of gender from the vari-ables used for adjustment, and we will be investi-gating a model-based method to solve this prob-lem.

Annual death rates & standard errorsFor this year’s ADR the biostatistical team imple-mented plans to use one mixed-effect Poisson re-gression model to obtain estimated death rates,and to use the bootstrap method to obtain accu-rate standard errors. We have performed exten-sive goodness-of-fit tests, and plan to furtherevaluate this model during the next year.

Standardized mortality, transplant, &hospitalization ratiosCalculation of SMRs, STRs, and SHRs is dis-cussed earlier in this Appendix. The currentmethod uses a Poisson model, adjusting for age,gender, race, and diabetic status, to calculate ex-pected mortality, transplant, and hospitalizationrates. The particular statistical model used tocompute expected values, as well as the variablesincluded in the model for comorbidity and se-verity of disease adjustment, can have large ef-fects on resulting ratios. During the next year, thebiostatistical team will investigate the strengthsand limitations of the current method.

Non-Medicare (‘ZZ’) patientsAs discussed at the beginning of this appendix,the difficulties of identifying non-Medicare (‘ZZ’)patients lead to problems in calculating the ac-tual number of patients with ESRD. Each non-Medicare patient is assigned a temporary Medi-care Beneficiary Claim number (HIC/BIC) with‘ZZ’ in the BIC field; this number is changed to apermanent HIC/BIC once the patient is eligiblefor Medicare entitlement. It is extremely impor-tant, therefore, for the USRDS to link all past ser-vices and new events so that records of these pa-tients’ medical and treatment history can beaccurately maintained and tracked. We plan tocollaborate with HCFA to create a consistentmethodology for reconciling the records of all


non-Medicare patients over the duration of theirtime on renal replacement therapy.

Dynamic web applicationThe USRDS currently maintains a Web site(www.usrds.org) with static pages of ESRD pa-tient information that can be viewed and down-loaded. To make this system more useful to re-searchers, and to allow users of the site immediateaccess to customized datasets, we are creating adynamic query application system. This interac-tive system will include a menu of data requestservices varying in complexity from overall pa-tient counts to individual rates by state and net-work, and will include the most recent informa-tion available on ESRD patients, updated on aquarterly basis. The initial version of the site willbe available by the summer of 2001, and we hopethat the completed site will become a fundamen-tal resource for publishing and sharing informa-tion on ESRD patients in the United States.

BIBLIOGRAPHYAmerican Diabetes Association. Standards ofmedical care for patients with diabetes mellitus.Diabetes Care 2000;23:S32-S42.

Bloembergen WE, Port FK, Mauger EA, WolfeRA. A comparison of mortality between patientstreated with hemodialysis and peritoneal dialy-sis. J Am Soc Nephrol 1995 Aug;6(2):177-83.

Collins AJ, Hao W, Xia H, Ebben JP, Everson SE,Constantini EG, Ma JZ. Mortality risks of peri-toneal dialysis and hemodialysis. Am J Kidney Dis1999 Dec;34(6):1065–74.

Collins AJ, Li S, Ebben J, Ma JZ, Manning W.Hematocrit levels and associated Medicare expen-ditures. Am J Kidney Dis 2000 Aug;36(2):282-93.

Fenton SS, Schaubel DE, Desmeules M, MorrisonHI, Mao Y, Copleston P, Jeffery JR, KjellstrandCM. Hemodialysis versus peritoneal dialysis: acomparison of adjusted mortality rates. Am JKidney Dis 1997 Sep;30(3):334–42.

Greenwood M. The natural duration of cancer. Re-ports on Public Health and Medical Subjects. Lon-don: Her Majesty’s Stationery Office 1926;33:1–26.

Herzog CA, Ma JZ, Collins AJ. Poor long-termsurvival after acute myocardial infarction amongpatients on long-term dialysis. N Engl J Med 1998Sep 17;339(12):799-805.

Hollenberg NK, Raij L. Angiotensin-convertingenzyme inhibition and renal protection: an as-

sessment of implications for therapy. Arch InternMed 1993;153:2426-2435.

Kalbfleisch JD, Prentice RL. The statistical analy-sis of failure time data. New York: Wiley, 1980.

Kaplan EL, Meier P. Nonparametric estimationfrom incomplete observation. J Amer Stat Assoc1958;3:457–481.

Mungoile M, Pickle LW, Simonson KH. Appli-cation of a weighted head-banging algorithm tomortality data maps. Statistics in Medicine1999;18,3201–3209.

National Committee for Quality Assurance.HEDISâ 2000 technical specifications. Washing-ton DC: National Committee for Quality Assur-ance; 1999.

Pei Y, Hercz G, Greenwood C, Segre G, ManuelA, Saiphoo C, Fenton S, Sherrard D. Risk factorsfor renal osteodystrophy: a multivariant analy-sis. J Bone Miner Res 1995 Jan;10(1):149-56.

Pickle LW, Mungiole M, Jones GK, White AA.Atlas of United States mortality. Hyattsville (MD):National Center for Health Statistics; 1996.

Pickle LW, White AA. Effects of the choice of age-adjustment method on maps of death rates. Sta-tistics in Medicine 1995;14,615–627.

SASÒ Institute Inc. SAS system for mixed models,Cary (NC): SAS Institute Inc.; 1996.

SASÒ Institute Inc. SAS/ETS User’s Guide. Cary(NC): SAS Institute Inc.; 1993.

Tukey PA, Tukey JW. Graphical display of datasets in 3 or more dimensions. In Barnett V, edi-tor. Interpreting multivariate data. New York:Wiley; 1981.

Visscher TL, Seidell JC. The public health impact ofobesity. Ann Rev Public Health 2001;22:355-375.

Vonesh EF, Moran J. Mortality in end-stage renaldisease: a reassessment of differences between pa-tients treated with hemodialysis and peritonealdialysis. J Am Soc Nephrol 1999 Feb;10(2):354–65.

Waller LA, Carlin BP, Xia H, Gelfand AE. Hierar-chical spatio-temporal mapping of disease rates.J Amer Stat Assoc 1997; 92,607–617.

A P P E N D I X B � E S R D N E T W O R K D E F I N I T I O N S � 2 3 1

Network 1Connecticut, Massachusetts, Maine, NewHampshire, Rhode Island, Vermont

Network 2New York

Network 3New Jersey, Puerto Rico, Virgin Islands

Network 4Delaware, Pennsylvania

Network 5Maryland, Virginia, Washington D.C.,West Virginia

Network 6Georgia, North Carolina, South Carolina

Network 7Florida

Network 8Alabama, Mississippi, Tennessee

Network 9Indiana, Kentucky, Ohio

Network 10Illinois

Network 11Michigan, Minnesota, North Dakota, SouthDakota, Wisconsin

Network 12Iowa, Kansas, Missouri, Nebraska

Network 13Arkansas, Louisiana, Oklahoma

Network 14Texas

Network 15Arizona, Colorado, Nevada, New Mexico, Utah,Wyoming

Network 16Alaska, Idaho, Montana, Oregon, Washington

Network 17American Samoa, northern California, Guam,Hawaii

Network 18Southern CaliforniaBecause of difficulties in identifying the networkof California patients, Networks 17 and 18 arecombined for the graphs included in this ADR,and are referred to as Network 17/18.

Appendix B · ESRD Network Definitions

Network 9

Network 1

Network 4 Network 3

Network 2

Network 5

Network 8

Network 6

Network 7

Network 14

Network 13Network 15

Network 12

Network 11Network 16

Network 17/18

Network 10


Appendix C · Glossary

Some of these definitions have been taken fromDorland’s Illustrated Medical Dictionary and fromthe On-line Medical Dictionary (http://www.graylab.ac.uk/omd/index.html).

ABO blood groupThe major human blood type system; importantin the determination of blood donors and bloodrecipients.

Acquired immunodeficiency syndrome (AIDS)An epidemic disease caused by the humanimmunodeficiency retrovirus that leads toimmune system failure, infections, and severeweakening of the body.

Adjusted average per capita costs (AAPCC )An estimate of how much Medicare will spendin a year for an average beneficiary.

Adult polycystic kidney disease (ADPKD)An inherited disease in which the kidneys containmultiple cysts; can cause chronic renal failure.

AngioplastyA procedure in which a balloon catheter isinserted into a blocked or narrowed vessel inorder to re-open the vessel and allow normal flowthrough it.

Atherosclerotic heart disease (ASHD)A disease of the arteries of the heart, characterizedby a thickening and/or loss of elasticity of thearterial walls.

Blood urea nitrogen (BUN)A by-product of the break-down of amino acidsand endogenous and injested protein.

Body mass index (BMI)A measure of height to weight ratio. BMI =Weight (kg) / Height (M2).

Conventional hemodialysisDialysis therapy using small surface areahemodialyzers that are made with conventionalmembranes and have low solute clearance andlow fluid removal capabilities. This type of dialysisdoes not require the use of delivery systems withultrafiltration control.

Coronary artery disease (CAD)A disease that causes narrowing or occlusion ofthe arteries surrounding the heart.

Continuous ambulatory peritoneal dialysis(CAPD)A type of dialysis in which dialysate is continu-ously present in the abdominal cavity. Fluid isexchanged by using gravity to fill and empty thecavity four to five times each day.

Continuous cycler-assisted peritoneal dialysis(CCPD)A type of dialysis in which the abdominal cavityis filled and emptied of dialysate using anautomated cycler machine.

CancerA disease that causes abnormal cell growth.

Cardiac arrestA complete cessation of cardiac activity.

CardiomyopathyA general diagnostic term indicating a primarynon-inflammatory disease of the heart muscle.

Cerebrovascular disease (CVD)A disease that causes narrowing or occlusion ofthe arteries supplying the brain. Cerebral vascularaccidents (CVA) and transient ischemic attacks(TIA) are two events that can result from thiscondition.

Common Working File (CWF) SystemThe Medicare Part A and Part B benefit coordi-nation and claims validation system. Under theCWF, HCFA maintains both institutional andphysician supplier claims-level data. CWF claimsrecords are the data source for most claims andutilization files used by the USRDS.

Congestive heart failure (CHF)A condition caused by ineffective pumping of theheart and subsequent fluid accumulation in thelungs.

Chronic obstructive pulmonary disease (COPD)A progressive disease characterized by coughing,wheezing, or difficulty in breathing.

Clinical Performance Measures ProjectFormerly the Core Indicator Project. A projectcooperatively managed by HCFA and the ESRDRenal Networks that maintains a clinical databaseof key elements related to the quality of dialysiscare. These elements are used as indicators inquality improvement initiatives.

A P P E N D I X C � G L O S S A R Y � 2 3 3

Health Care Financing Administration (HCFA)Federal agency that administers the Medicare,Medicaid, and State Childrens’ Health insuranceprograms.

Health Plan Employee Data Information Set(HEDISâ 3.0)Established by the National Committee forQuality Assurance, HEDISâ 3.0 is a set ofstandardized performance measures establishedto aid consumers in the comparison of managedhealthcare plans.

HepatitisAn inflammation of the liver that may be causedby a viral infection, poisons, or the use of alcoholor other drugs. The disease takes on many formswhich include Hepatitis A, a form of the virususually transmitted by contaminated food orwater; Hepatitis B, more serious than HepatitisA and transmitted through blood and bodyfluids; Hepatitis C, also transmitted throughblood and body fluids; and Hepatitis D, whichcauses symptoms only when an individual isalready infected with the Hepatitis B virus.

High-efficiency dialysisDialysis therapy that is provided usinghemodialyzers with larger surface areas thanconventional hemodialyzers. Enhanced soluteclearance is achieved through increased bloodflow rates of 300 to 400 milliliters per minute,allowing treatment times to be reduced toapproximately three hours.

High-flux dialysisDialysis therapy provided using hemodialyzerswith synthetic membranes and large surface areasthat, combined with high blood and dialysate flowrates, allow enhanced solute clearance and fluidremoval. Delivery systems with ultrafiltrationcontrol are required for this therapy.

Hospital center unitA dialysis unit located in or attached to a hospitaland licensed to furnish inpatient and outpatientdialysis plus diagnostic, therapeutic and rehabili-tative services.

Hospital facility unitA dialysis unit attached to or located in a hospitaland licensed to provide outpatient dialysisservices directly to ESRD patients.

Incident patientA patient starting renal replacement therapy forend-stage renal disease during the calendar year.This definition excludes persons with acute renal

CreatinineA waste product of protein metabolism found inthe urine and often used to evaluate kidneyfunction. Abnormally high creatinine levels areseen in individuals with kidney failure or kidneyinsufficiency.

Creatinine clearanceUsed as an indicator to predict the onset ofuremia, which develops when the creatinineclearance falls below 10 ml/minute/1.73M2.

Death Notification Form (HCFA-2746)A form submitted following the death of anESRD patient, and containing basic patientdemographic information in addition toinformation on the primary cause of death.

Diabetes mellitus, insulin dependentA condition in which insulin is necessary toregulate abnormally high levels of glucose (sugar)in the blood.

Diagnosis Related Groups (DRGs)Used by Medicare to determine payment forinpatient hospital stays; based on diagnosis, age,gender, and complications.

Dialysis Outcomes Quality Initiative (DOQI)Established in 1995 by the National KidneyFoundation to improve patient outcomes andsurvival by providing recommendations foroptimal clinical practices in the areas of dialysisadequacy, vascular access, and anemia.

Dialysis and transplant centerA facility that combines the functions of a dialysiscenter and a transplant center.

End-stage renal disease (ESRD)A condition in which an individual’s kidneyfunction is not adequate to support life.

Erythropoietin (EPO)A hormone secreted chiefly by the adult kidney;acts on the bone marrow to stimulate red cellproduction.

Freestanding unitA dialysis unit licensed to provide only outpatientand home maintenance dialysis; sometimesreferred to as an independent unit.

Glomerular filtration rate (GFR)The rate at which the kidneys remove waste prod-ucts from the blood.


failure, persons with chronic renal failure who diebefore receiving treatment for ESRD, and personswhose ESRD treatments are not reported throughHCFA.

Ischemic heart disease (ISHD)A disease of the heart evidenced by a loweredoxygen supply to the heart tissue caused byocclusion or narrowing of the arteries supplyingthe heart muscle.

Kt/VAn indicator of the dose of dialysis received pertreatment. Dose is calculated by multiplying theurea clearance (K) times the treatment duration(t) and dividing by the urea distribution (V). Theurea distribution volume is approximately equalto the volume of total body water.

Medical Evidence Form (HCFA-2728)A form which provides source data about ESRDpatients, including information on patientdemographics, primary cause of renal disease,comorbidity, laboratory values, dialysis treatment,transplant, dialysis training, employment status,and initial insurance coverage.

Myocardial infarction (MI)An event which causes injury to the heart muscle,also called a heart attack.

National Claims History (NCH) 100% NearlineFileA file which contains all Common Working File(CWF) Part A (provider) and Part B (physician/supplier) Medicare claims and adjusted claimsinformation.

Period prevalent patientA patient receiving treatment for ESRD at somepoint during a period of time, usually six monthsor a year. Patients may die during the period orbe point prevalent at the end of the period.

Peripheral vascular disease (PVD)A progressive disease that causes narrowing orocclusion of the arteries supplying the extremitiesof the body.

Peritoneal dialysisA type of dialysis in which fluid (dialysate) isintroduced into the abdominal cavity and uremictoxins are removed across the peritoneum.

Point prevalent patientA patient reported as receiving treatment forESRD on a particular day of the calendar year(i.e. December 31st).

Program Medical Management and InformationSystem for ESRD and Renal Beneficiary and Uti-lization System /Program (PMMIS/REBUS)The major source of data for the USRDS. ThisHCFA file incorporates data from the MedicalEvidence Form (HCFA-2728), the DeathNotification Form (HCFA-2746), the MedicareEnrollment Database, HCFA Paid ClaimsRecords, and the UNOS Transplant Database.

Prevalent patientA patient receiving renal replacement therapy orhaving a functioning kidney transplant(regardless of when the transplant wasperformed). This definition excludes personswith acute renal failure, persons with chronicrenal failure who die before receiving treatmentfor ESRD, and persons whose ESRD treatmentsare not reported through HCFA.

PyrogenA substance which is bacterial in nature andcapable of producing low grade fevers.

Pyrogen reactionA condition in which a patient who was afebrileprior to dialysis experiences a low-grade feverduring dialysis, caused by pyrogens in thedialysate fluid. The fever disappears after dialysisis over, distinguishing the reaction from an actualinfection.

REMIS/PMMISHCFA’s Renal Management InformationSystem (REMIS)/Program Management andMedical Information System (PMMIS) iscurrently under development and isanticipated to replace the existing RenalBeneficiary and Utilization System (REBUS/PMMIS) in the summer of 2001. The firstrelease, which will incorporate most of thecapabilities, interfaces, and processes of thecurrent system, will further support andimprove data collection, validation, andanalysis, and will provide timely and accurateinformation to the ESRD Networks, dialysisfacilities, transplant centers, and researchorganizations. This new system willsignificantly improve support for ESRDprogram analysis, policy development, andepidemiological research.

Renal networkEstablished in 1978 as a provider oversightsystem that assures ESRD patients are providedimmediate access to treatment and that the carethey receive meets the highest qualitystandards.

A P P E N D I X C � G L O S S A R Y � 2 3 5

ReuseA process through which a hemodialyzer iscleaned and disinfected, allowing it to be usedmultiple times on the same patient.

Reuse germicideA chemical used during the reuse process todisinfect the hemodialyzer.

SIMSHCFA’s Standard Information ManagementSystem (SIMS), which became operational atthe beginning of 2000. Supports the HCFA re-porting requirements and business processesof the ESRD Networks; provides communica-tion and data exchange links among the Net-works, HCFA and other segments of the renalcommunity to support quality improvementactivities relating to the treatment of ESRD;supplies standard core data functionality forprevious Network data systems; and providesimproved electronic communication capabili-ties, data standardization, and informationmanagement tools.

Standard Analytical Files (SAFs)HCFA files which contain final action MedicarePart A claims data. SAFs are comprised of eightfiles: Inpatient, Outpatient, Home Health Agency,Hospice, Skilled Nursing Facility, ClinicalLaboratory, Durable Medical Equipment, and 5%Sample Beneficiary.

Standardized hospitalization ratio (SHR) selected group of patients by computing the ratio

of the group’s observed hospitalization rate to theexpected hospitalization rate for the nationalESRD population.

Standardized mortality ratio ( SMR )Used to compare dialysis patient mortality ratesfrom year to year. Mortality rates for a subgroupof patients are compared to a set of reference rates,with adjustments for age, race, gender, anddiabetes as a cause of ESRD.

Standardized transplantation ratio (STR)Used to compare the transplant rate of a sub-group of patients to the national transplant rate.

Transplant centerA hospital unit licensed to provide transplanta-tion and other medical and surgical specialtyservices for the care of kidney transplant patients,including inpatient dialysis furnished directly orunder arrangement.

United Network for Organ Sharing (UNOS)A private, non-profit organization that maintainsthe organ transplant list for the nation andcoordinates the matching and distribution oforgans to patients awaiting transplant.

Urea reduction ration (URR)A means of measuring dialysis dose by calculatingthe change in BUN over the course of a dialysistreatment. URR = (pre-dialysis – post-dialysisBUN) / pre-dialysis BUN * 100.

Valvular heart disease (VHD)A condition in which a patient has one or moreabnormal heart valves.

VintageTime in years that a patient has had ESRD.

The VISION projectHCFA’s Vital Information System to ImproveOutcomes in Nephrology (VISION) will providecustomized data entry and reporting for thenearly 4,000 U.S. dialysis facilities, and will captureand securely communicate ESRD patient andprovider data collected via the HCFA-2728,HCFA-2746, HCFA-2744, HCFA-820 andHCFA-821 forms for subsequent electronicreporting to the ESRD Network Organizationsand HCFA. This project is designed to meet thegoals of the Hemodialysis Facilities ofAchievement Project (FOA) as outlined in theFederal Register (April 29, 1997) and is furthermandated by the Balanced Budget Act (BBA) of1997 to be implemented by the beginning of theyear 2001.


NATIONAL INSTITUTE OF DIABETES AND DIGESTIVEAND KIDNEY DISEASES (NIDDK)Josephine Briggs, MDDirector, Division of Kidney, Urologic, andHematologic Diseases (DKUHD)

Lawrence Y.C. Agodoa, MDCo-Project Officer, USRDS;Director, End-Stage Renal Disease Program,DKUHD

Paul W. Eggers, PhDCo-Project Officer, USRDS;Program Director, Kidney and UrologyEpidemiology, NIDDK

Robert CoonleyContracting Officer, Contracts ManagementBranch, NIH

Marlene MirelesContract Specialist, Cardiovascular SpecialStudies Center

HEALTH CARE FINANCING ADMINISTRATION(HCFA)Pamela R. Frederick, MSBHCFA Project Coordinator, USRDS; TechnicalAdvisor for the ESRD Program, Quality Measure-ment and Health Assessment Group, Office ofClinical Standards and Quality (OCSQ)

Joel W. Greer, PhDSocial Science Research Analyst, Office of StrategicPlanning

Jeffrey L. Kang, MD, MPHChief Medical Officer and Director, OCSQ

Roger MilamComputer Specialist, Division of ESRD Systemsand Contract Management, Information SystemGroup, OCSQ

Ida SarsitisDirector, Division of Contract Policy and Perfor-mance, Quality Improvement Group, OCSQ

Dennis StrickerActing Director, Information System Group,OCSQ

USRDS COORDINATING CENTER (CC),MINNEAPOLIS MEDICAL RESEARCH FOUNDATIONAllan J. Collins, MD, FACPDirector; Professor of Medicine, University ofMinnesota School of Medicine; Nephrologist,Department of Medicine, Hennepin CountyMedical Center

Bertram Kasiske, MDDeputy Director; Professor of Medicine, Univer-sity of Minnesota School of Medicine; Chief ofNephrology, Department of Medicine, Henne-pin County Medical Center

Delaney Berrini, BSAssistant Editor, Administrative Assistant

Shu Chen, MSDirector of Information Systems and SoftwareDevelopment

Edward Constantini, MAAssociate Editor, Research Associate

Frederick Dalleska, MSSenior Systems Analyst/Programmer

Stephan Dunning, BAGeographic Information Systems

James Ebben, BSManager, Data Systems

Susan Everson, PhDSenior Editor, Research Associate

Beth Forrest, BBAAdministrative Assistant

Eric Frazier, BSSystems Analyst/Programmer, Webmaster

David Gilbertson, PhDBiostatistician

Dana Knopic, AASAdministrative Manager

Roger JohnsonSenior Systems Analyst/Programmer

Shuling Li, MSBiostatistician

Appendix D · USRDS Staff & Co-investigators

A P P E N D I X D � U S R D S S T A F F & M E M B E R S H I P � 2 3 7

Suying Li, MSBiostatistician, Economist

Jiannong Liu, MSBiostatistician

Tricia Roberts, MSBiostatistician

C. Daniel Sheets, BSSenior Systems Analyst/Programmer

Jon Snyder, MSBiostatistician

Jay Xue, DVM, PhDEpidemiologist

USRDS CO-INVESTIGATORSBlanche Chavers, MDAssociate Professor of Pediatrics, University ofMinnesota School of Medicine

Richard Grimm, MD, PhDProfessor of Medicine and Epidemiology,University of Minnesota School of Public Health;Director of Clinical Epidemiology, Departmentof Medicine, Hennepin County Medical Center

Charles Herzog, MDStaff Cardiologist, Department of Medicine,Hennepin County Medical Center; AssociateProfessor of Medicine, University of MinnesotaSchool of Medicine

William Keane, MDProfessor of Medicine, University of MinnesotaSchool of Medicine; Chairman, Department ofMedicine, Hennepin County Medical Center

Thomas Louis, PhDRand Corporation

Jennie Z. Ma, PhDAssistant Professor, Department of PreventiveMedicine, University of Tennessee College ofMedicine

Willard Manning, PhDProfessor, Department of Health Studies, HarrisSchool of Public Policy Studies, University ofChicago

Arthur Matas, MDProfessor, Department of Surgery, University ofMinnesota School of Medicine

Marshall McBean, MD, MScProfessor and Department Head, Department ofHealth Management and Policy, University ofMinnesota School of Public Health

Wei Pan, PhDAssistant Professor, Division of Biostatistics,University of Minnesota School of Public Health

Wendy L. St. Peter, PharmD, FCCP, BCPSAssociate Professor, University of MinnesotaCollege of Pharmacy; Department of Medicine,Hennepin County Medical Center

CARDIOVASCULAR SPECIAL STUDIES CENTERCharles Herzog, MDDirector; Staff Cardiologist, Department ofMedicine, Hennepin County Medical Center;Associate Professor of Medicine, University ofMinnesota School of Medicine

ECONOMIC SPECIAL STUDIES CENTERLawrence Hunsicker, MDDirector; Professor of Internal Medicine, Univer-sity of Iowa College of Medicine


Appendix E · USRDS Products & Services

Table e.1 describes the products and services pro-vided by the USRDS to support ESRD researchand the work of the renal community.

The entire ADR is available on the Internet atwww.usrds.org; included on the site as well arecolor slides of figures, a PDF file of theResearcher’s Guide, and USRDS contact infor-mation. In the future the site will allow users tocreate customized data sets and regional maps.Data on site use are presented in Figure e.1.

Dialysis unit-specific SMR/SHR reportsFrom 1996 through 1999 the USRDS producedmore than 2,300 unit-specific reports each year,compiling information about the patients treatedin each dialysis facility, and including Standard-ized Mortality Ratios (SMRs) and StandardizedHospitalization Ratios (SHRs). These reports arenow being produced by the Kidney Epidemiol-ogy and Cost Center at the University of Michi-gan (www.med.umich.edu/kidney).

SMR/SHR spreadsheetsThe USRDS produces Standardized MortalityRatio (SMR) and Standardized HospitalizationRatio (SHR) spreadsheets, available upon request.These spreadsheets allow comparison of thenumber of deaths or first hospitalizations for aspecific subgroup of patients (e.g., the patientsfor a single dialysis facility) to national rates basedon the actual mortality and hospitalization of U.S.dialysis patients. The SMRs and SHRs computedfrom these spreadsheets, however, are not directly

comparable to those provided on the dialysisunit-specific mortality and hospitalizationreports. The USRDS produces the national ratesused to compute the expected number of deathsand hospitalizations in the spreadsheet, while theKidney Epidemiology and Cost Center at theUniversity of Michigan independently calculatesthe national rates used for the SMRs and SHRsin the unit-specific reports, and the two groupsuse REBUS files cut at different times. To requestthe USRDS spreadsheets, contact the USRDSCoordinating Center at [email protected].

Data requestsMaking information on ESRD available to therenal community is a primary objective of theUSRDS, and we are committed to the timelyfulfillment of data requests. In many cases theserequests can be answered by providing datapublished in the ADR or elsewhere. Requests fordata not available in material published by theUSRDS, but that require two hours or less of stafftime, are fulfilled by the Coordinating Centerwithout charge, usually within one week. Morecomplex requests—those requiring more thantwo hours of staff time—as well as requests forStandard Analysis Files and custom files, mustbe accompanied by a written proposal (see detailsbelow), and will be completed only upon writtenapproval by the NIDDK Project Officer.

DATA FILES AVAILABLE TO RESEARCHERSThe Coordinating Center maintains a set ofStandard Analysis Files (SAFs) to meet diverse

Figure e.1Data requests & visits to theUSRDS website

Dec 99 Mar 00 Jun 00 Sep 00 Dec 00 Mar 010

10

20

30

40

50

60

70Data requests Website visits

Jul 00 Sep 00 Nov 00 Jan 01 Mar 01

500

1,000

1,500

2,000

A P P E N D I X E � U S R D S P R O D U C T S & S E R V I C E S � 2 3 9

Table e.1USRDS products & servicesfor ESRD researchers & thegeneral renal community

Products are provided withoutcharge except as noted.

research needs and to provide easy access to thedata used in the ADR. The SAFs were introducedin 1994, and at the same time NIDDK beganawarding a new group of grants focusing onresearch using the USRDS data. The result hasbeen an annual increase in the number of filesprovided by the USRDS.

Prior to 1994 all files provided to researchers werecustom files created for a specific research project.Since the introduction of the SAFs, however,custom files are generally limited to cases in whicha researcher provides a patient finder file to bematched with the USRDS database.

The Core SAF CD-ROM contains basic patientdata and is needed in order to use any of the otherSAFs. Included on this CD are each patient’sdemographic information, treatment history,limited transplant data, and all data from theUSRDS Special Studies. Approximately half of theresearchers using the USRDS SAFs need only this

CD. Full transplant information is provided on aseparate CD that contains detailed transplant andtransplant follow-up data collected by HCFA andUNOS. Data on hospital inpatient stays is foundon the hospitalization CD, and Medicare paymentdata is available either in a full set or by individualyear. See Table e.2.

STANDARD ANALYSIS FILES (SAFs)The use of the SAFs is governed by the USRDS“Policy on Data Release for Investigator-InitiatedResearch,” which appears later in this appendix.A researcher’s proposal must be approved by theUSRDS Project Officer, and the researcher mustsign the USRDS “Agreement for Release of Data”(see last page of appendix). Prices for these filesare listed in Table e.3.

Most SAFs provide patient-specific data. Allpatient identifiers (name, address, Social Securitynumber, etc.) are removed from the files orencrypted, but confidentiality of the data is still a

USRDS Coordinating Center914 South 8th Street, Suite D-206Minneapolis, MN 55404

Reports & guidesAnnual Data Reports Available from the National Kidney and Urologic Diseases Information Clearinghouse, 3

Information Way, Bethesda, MD 20892-3560, 301.654.4415, [email protected].

Material from the ADR will also be published in the American Journal of Kidney Disease.

Researcher’s Guide to the USRDS

Database

Provides a detailed description of the USRDS database and of the USRDS Standard Analysis

Files, and is the basic reference for researchers who use USRDS data files.

ADR CD-ROM Contains the text and graphics of the ADR, supplementary data tables, color Powerpoint

slides, and The Researcher’s Guide .

Internet site: www.usrds.org Contains PDF files of the chapters, reference tables, and Researcher’s Guide; Powerpoint

slides of Atlas figures and USRDS conference presentations; ASCII files of the data tables;

notices regarding current news and analyses; links to related Internet sites; and email

addresses for contacting the USRDS.

Requests for dataData requests: two-hour Questions and data requests that are not answered directly by the ADR can be addressed to

the Coordinating Center; those that require less than two hours of staff time to fulfill will be

processed without charge.

Data requests: more than two hours Those questions and data requests that require over two hours of staff time will have to be

reviewed and approved by the NIDDK Project Officer. Costs for this type of request will be

assessed on a case-by-case basis.

Standard Analysis Files (SAFs) These data files provide patient-specific data from the USRDS database to support ESRD

research. A standard price list has been established for the SAFs (table f.3), and users must

sign a data release agreement with NIDDK.

Custom data files Custom files can be created by the Coordinating Center for projects requiring data other

than that provided in the Standard Analysis Files. An hourly rate of $72.70 will be assessed

for time spent on the request, and users must sign a data release agreement with NIDDK.

Papers, abstracts, & publicationsMost USRDS research studies result in published papers or presentations at professional

meetings. Figures from presentations can be found on the website, while published

abstracts and papers can be found in the relevant journals.

Contact informationData requests & publication orders 612.347.7776 or 1.888.99USRDS

Fax 612.347.5878 www.urrds.orgData file contact Shu Chen, MS, [email protected]


serious concern. The “Agreement for Release ofData” therefore includes restrictions on the useand disposition of the SAFs. The SAFs do includean encrypted ID number to allow patient datafrom multiple SAFs to be merged when needed.

Core Standard Analysis File CD-ROMThe USRDS has carried out a number of SpecialStudies. Topics are approved by the NIDDK, withrecommendations from HCFA, the USRDS Sci-entific Advisory Committee, the ESRD networks,and the Renal Community Council. For eachstudy, design and sampling plans were developed,samples were selected, and data collection formsand instructions were drafted, tested, and final-ized. The main studies are summarized below anddetailed in the Researcher’s Guide.

This CD contains the most frequently usedSAFs, including those from the USRDS Spe-cial Studies, and is needed for use of the Trans-plant CD, the Hospital CD, or any CD basedon Medicare claims data. The files included onthis Core CD are as follows (and are also listedin Table e.2):

PATIENT

Contains one record per patient in the USRDSdatabase, and gives basic demographic andESRD-related data.

RESIDENCE

Provides a longitudinal record, to ZIP code level,of each patient’s place of residence.

TREATMENT HISTORY

Also referred to as the Modality Sequence file;contains a new record for each patient at eachchange in treatment modality or dialysis provider.

MEDICAL EVIDENCE

Contains full data from the 1995 version of theHCFA Chronic Renal Disease Medical EvidenceForm (HCFA-2728), the source of data about theprimary disease causing renal failure and the startdate of chronic renal dialysis. In April 1995 a newversion of the form went into use that includeddata on comorbidity, employment status, lab val-ues at start of dialysis, and Hispanic ethnicity.

TRANSPLANT

Contains basic data for all transplants, includinggraft failure date (detailed transplant data are con-tained on a separate transplant CD).

TRANSPLANT WAITING LIST

Includes one record for each patient in theUSRDS database who also can be identified in

the UNOS transplant waiting list file, and con-tains only the date on which the patient wasfirst placed on the waiting list. Because of thecomplexity and variability of the patterns ofpatient movement on and off the waiting list,we have not attempted to derive more com-plex indicators of transplant waiting list expe-rience.

DIALYSIS MORBIDITY & MORTALITY STUDY

The DMMS was an observational study in whichdata on demographics, comorbidity, laboratoryvalues, treatment, socioeconomic factors, andinsurance were collected for a random sample ofU.S. dialysis patients, using dialysis records. Datawas collected on 6,000 ESRD patients in each ofWaves 1, 3, and 4 and 4,500 patients in Wave 2, atotal of 22,500 patients over three years. Waves 1,3, and 4 are each historical prospective studies inwhich data were collected for patients receivingin-center hemodialysis on December 31, 1993.Data were abstracted from the patient’s medicalrecord and the patient was followed from Decem-ber 31, 1993 through the earliest of data abstrac-tion, death, transplant, change in modality, ortransfer to another facility. Wave 2 is a true pro-spective study of incident hemodialysis and peri-toneal dialysis patients for 1996.

CASE MIX ADEQUACY STUDY

The objectives of the USRDS Case Mix AdequacyStudy of Dialysis were to

¨ establish the relationship between the doseof delivered dialysis therapy and mortal-ity

¨ determine the strength of this relationshipwhen data are adjusted for comorbidity

¨ assess how this relationship changes at dif-ferent dialysis doses

¨ assess how this relationship is affected bydialyzer reuse

¨ assess the impact of different dialysismembranes on patient morbidity andmortality

The study consisted of two groups of patients:an incident sample of ESRD patients who be-gan hemodialysis during 1990, and a prevalentsample of hemodialysis patients with onset ofESRD prior to 1990. A total of 7,096 patientsfrom 523 dialysis units were included, withapproximately 3,300 patients having the pre-and post- BUN values needed to calculate de-livered dialysis dose. Ninety-four percent ofthese cases were matched to the USRDS data-base. The ESRD Networks collected these datain conjunction with their Medical Case Reviewdata abstraction.


Table e.2Contents of the USRDS CoreStandard Analysis File CD-ROM

This file is needed in order touse any of the other StandardAnalysis Files. The data are pro-vided on a single CD-ROM.

CASE MIX SEVERITY STUDY

The objectives of this study were to:¨ estimate the correlation of comorbidity

and other factors existing at the onset ofESRD to subsequent mortality and hos-pitalization rates, while adjusting for age,gender, race, and primary diagnosis

¨ evaluate possible associations of thesefactors with reported causes of death

¨ assess the distribution of comorbidityand other factors among patients usingdiffernet treatment modalities

¨ compare relative mortality rates by treat-ment modality, adjusting for selectedcomorbid conditions and other factors

Data were collected on 5,255 patients incident in1986–87 at 328 dialysis units nationwide.

PEDIATRIC GROWTH & DEVELOPMENT

The objectives of the USRDS Pediatric ESRDGrowth and Development Study were to:

¨ establish a baseline for assessing the rela-tion of pediatric ESRD patient growth andsexual maturation to modality

¨ establish a prototype for the ongoing col-lection of pediatric data

All patients who were prevalent in 1990 andborn after December 31, 1970 were includedin the study, a total of 3,067 patients at 548dialysis units.

CAPD & PERITONITIS STUDY

The USRDS CAPD and Peritonitis Rates Studyexamined the relation of peritonitis episodesin CAPD patients to connection devicetechnology and other factors. The studypopulation included all patients newly startingCAPD in the first six months of 1989, up to amaximum of 14 patients per dialysis unit. Allunits providing CAPD training participated inthe study. The sample contains 3,385 patientsfrom 706 units.

File Name Unit of observation UsesPatient One record for each ESRD

patient.

Incidence, prevalence, patient survival. Most other files will

need to be linked to this file using the encrypted patient ID.

Residence For each patient, one record for

each period in a different

residence.

Regional analyses.

Treatment History One record for each period a

patient is on one modality.

Modality distribution and treatment patterns. Treatment

modality at a point in time and modality changes over time.

Medical Evidence One record for each 2728 form

filed (1995 version).

ESRD first service date, initial treatment modality, comorbid

conditions, patient status at start of ESRD.

Transplant One record for each transplant

event. Can have multiple

transplants for one patient.

Transplant and transplant outcome analyses.

Transplant Waiting List One record for each patient

ever on waiting list. Only data

item is date first listed.

Comparison of transplanted patients to dialysis patients who

are transplant candidates. Patient selection to waiting list.

Dialysis Mortality and

Morbidity (DMMS) (Special

Study)

Wave 1: 5,670 patients;

Wave 2: 4,024 patients;

Wave 3: 11,142 patients.

Comorbid conditions, adequacy of dialysis, dialysis

prescription and other treatment parameters, laboratory test

values, nutrition, vascular access. Case Mix Adequacy (Special

Study)

7,096 patients. Comorbid conditions, adequacy of dialysis, dialysis pre-

scription and other treatment parameters, laboratory values.

Case Mix Severity (Special

Study)

5,255 patients. Comorbid conditions, adequacy of dialysis, dialysis pre-

scription and other treatment parameters, laboratory values.

Pediatric Growth and

Development (Special Study)

3,067 patients. Growth, development, and other issues relating to pediatric

ESRD Patients.

CAPD Peritonitis (Special

Study)

3,385 patients. CAPD and peritonitis.

Facility One record for each year facility

operated.

Merge with the treatment history, transplant, or annual

summary SAFs for analyses involving provider characteristics

by encrypted ID.Facility Cost Reports One record per facility per year

(1989-1995 only).

Costs and staffing of dialysis facilities.

Dialyzers Information on dialyzer

characteristics; to be matched

to patient dialyzer information

in other files on CD

Relation of dialyzer characteristics to patient outcomes.

CLMCODES One record for each diagnosis,

procedure, or HCPCS code

appearing in claims files.

Frequency of occurrence of each code. A starting point for

analyses that will use diagnosis and procedure codes.

FORMATS.SC2 All USRDS-defined SAS formats

used by the SAFs.

This is a SAS format library to format values of categorical

variables.


FACILITY

The HCFA ESRD Annual Facility Survey is thesource of data for the Facility SAF, which can belinked to the Facility Cost Report files using theUSRDS provider ID. Because of this link,geographic variables that could be used to identifyfacilities have been deleted. The survey period isJanuary 1 through December 31.

FACILITY COST REPORTS

The HCFA hospital and independent facilitycost reports for the years 1989–1995 and 1989–1993 respectively are available as StandardAnalysis Files. All geographic variables havebeen deleted in order to ensure confidentiality.The file may be linked with the Facility SAF byusing the USRDS provider ID; geographicanalyses at less than a regional or ESRDnetwork level, however, are not possible.Because there has been minimal use of thesefiles, data for additional years will be added onlyif there is sufficient demand.

DIALYZERS

The Case Mix Severity, Case Mix Adequacy, andDMMS Special Studies all collected informationon the manufacturer and model of the dialyzerused for a patient at a specific time. The SAFs forthese studies describe the dialyzer only through acode, which must be matched to information inthe Dialyzer file to find the manufacturer andmodel of the dialyzer along with characteristicssuch as membrane type and clearance. The datain this file come from published sources avail-able at the time of the study. We believe these dataaccurately represent the dialyzer characteristics,but they should be used with caution.

Transplant CDDue to changes in data collection sources overthe years, data pertaining to transplants are nowpresented in six separate SAFs. The first two filesare included on the Core CD, and the remainingfour are included on the separate Transplant CD.

¨ TX: includes minimum details about alltransplants from all sources

¨ TXWAIT: contains one record for eachpatient in the USRDS database who alsocan be identified on the kidney transplantwaiting list maintained by the United Net-work for Organ Sharing (UNOS); the onlyvariables are the date of first listing andUSRDS_ID

¨ TXHCFA: includes transplant detailscollected by HCFA’s PMMIS system priorto 1994

¨ TXUNOS: includes transplant detailscollected by UNOS, currently the main

source of transplant data for the USRDS,for the years since 1987

¨ TXFUHCFA: includes transplant follow-up reports collected by HCFA prior to1994. Reports are completed at discharge,six months, each year post-transplant, andgraft failure

¨ TXFUUNOS: includes transplantfollow-up reports collected by UNOSsince 1987.

The tables in Section F of the Reference Tablesare produced primarily from the main and UNOStransplant files.

In July 1994 HCFA and the Health ResourcesServices Administration (HRSA) consolidatedtransplant data into a single collection by UNOSunder its contract with HRSA. The expandedtransplant data are shared among HRSA, HCFA,and the NIH and thus are available to the USRDS.This has resulted in the addition of data on a sub-stantial number of non-Medicare transplantpatients, including children.

The HCFA and UNOS transplant data files over-lap for 1987–1993, and some Medical EvidenceForms and institutional claims records indicatetransplants that are not included in either theHCFA or the UNOS file. As reported in Appen-dix A, the following procedure is used to resolvethe conflicts among the four sources and createthe transplant SAF. All UNOS transplants are firstaccepted into the file, with all HCFA transplantsprior to 1987 accepted next. HCFA transplantsfrom 1987–1993 are then accepted if there is notransplant in the file for that patient within 30days of the HCFA transplant (it is common forthe transplant dates to differ by one day betweenthese two sources). Finally, transplants indicatedon the Medical Evidence 2728 Form are acceptedif no transplant is listed for that patient within 30days of the Medical Evidence transplant date.

Hospital CDHospitalization inpatient data from the USRDSdatabase are a subset of the data in the Institu-tional Claims file. No payment or cost variablesare included on this CD. This CD is for research-ers who need data on hospital inpatient stays andon diagnoses and procedures for those stays butwho do not need payment data.

Dialysis Morbidity & Mortality CDThis CD contains files from the Dialysis Morbid-ity and Mortality Study and extracts data fromall other SAFs for the patients in this study. Alldata on Medicare payments for these patients arefollowed to the currently reported claims year.


Case Mix Adequacy CDThis CD contains the Case Mix Adequacy SpecialStudy file and extracts data from all other SAFsfor the patients in this study. All data on Medicarepayments for these patients are followed to thecurrently reported claims year. Along withanalyses related to the study itself, this file is usefulfor developing analyses that will later be run onthe full Medicare payments files.

CDs of Medicare payment dataMedicare payment data on institutional claimsare available for pre-1989 through 1999, whiledata on physician/supplier claims are available for1991–1999. The 1999 claims will be availablealong with other USRDS SAF CDs in summer2001. These data sets can be purchased by year.

Institutional claims consist of all Part A claims(Inpatient, Outpatient, Skilled Nursing Facility,Home Health Agency, and Hospice) and somePart B claims, notably outpatient dialysis. All phy-sician/supplier claims are Medicare Part B; theseclaims account for about 80 percent of the claimsbut only 20 percent of the dollars.

The structure and content of the two types ofclaims are different, as are the files derived from

them. Institutional claims are provided in two filetypes: the Institutional Claims file, whichindicates the type of claim, the dollar amounts,the DRG code, the type of dialysis involved (ifany), and the dates of service; and the InstitutionalClaims Detail file, which contains details such asdiagnosis and procedure codes. Many analyseswill require only the Institutional Claims files.

Physician/supplier claims are contained in onetype of file with one record for each claim line-item. The file includes dollar amounts, dates ofservice, diagnosis and procedure codes, and typeand place of service.

File media and formatsThe SAFs are provided on CD-ROM disks asSASÒ (Statistical Analysis System) files, and canbe used directly by SASÒ on any 486 or PentiumPC with a CD-ROM reader.

A SASÒ format was chosen for the USRDS SAFsbecause it is widely used, easily transported, andlargely self-documenting. SASÒ is a commerciallyavailable data management and statistical analysissoftware system that runs on most computers,from mainframes to PCs, and it is almostuniversally available on university computer

Table e.3Prices for USRDS StandardAnalysis Fileson CD-ROM

Standard Analytical File CD-ROMsCD Description CDs PriceCore CD The Core CD is needed in order to use any of the other CDs. 1 $500Transplant CD Detailed transplant data from HCFA and UNOS. 1 $100

Hospital CD The Hospital CD is derived from the Institutional Claims and Institutional 1 $100Claims Details CDs. It contains diagnosis and surgical procedure codes for eachstay but does not include the cost data from the Institutional Claims records.

DMMS claims CD set The DMMS claims CD set contains all of the Institutional and 3 $300claims data for the patients in the USRDS Dialysis Morbidity and Mortality Special Study. The data from the Special Study data collection forms are on the Core CD.

Case Mix Adequacy The Case Mix Adequacy Claims CD set contains all of the Institutional and 1 $100claims CD Physician/Supplier claims data for the patients in the USRDS Case Mix

Special Study. The data from the Special Study data collection forms are included.

Medicare payment data

Years Claims CDs Details CDs Price CDs PriceBefore 1989* 1 1 $2001989 1 1 $2001990 1 1 $2001991 1 2 $300 4 $4001992 1 2 $300 4 $4001993 1 2 $300 4 $4001994 1 2 $300 5 $5001995 1 3 $400 5 $5001996 1 3 $400 6 $6001997 1 3 $400 7 $7001998 1 3 $400 7 $7001999***CDs for years prior to 1989 included only hospital inpatient stays and quarterly summaries of outpatient dialysis; no cost data is included.**The number of CDs for the Institutional claims and Physician/Supplier claims have not yetbeen determined. The total number of CDs, however, will be at least the same as in 1998,if not more. The unit price will remain the same.

Institutional Claims Physician/Supplier Claims


I Research topic title and submission date

II Background information

III Study designObjectivesHypothesisAnalytical methods

IV Data being requestedList Standard Analytical Files needed, or specify fields

needed in custom data fileDescribe data security: responsible party, computer access,

etc.

V Investigator informationFor Principal Investigator and co-authors, supply:

NameAffiliationAddressPhone numberFax numberEmail address

Table e.4Suggested outline for researchproposals using USRDS data

systems. The USRDS SAFs take full advantage ofthe program’s ability to incorporate a largeamount of documentation into the file.

Researchers who require a different programformat or a medium other than CD-ROM willneed to arrange for the conversion themselves.The USRDS also may be able to convert files toalternative formats or media, but the cost will besubstantially greater.

What is needed to use the SAFs¨ Computer: at a minimum, a 486 or

Pentium PC. Smaller runs have been doneon 486/100 PCs. The files can be convertedto SASÒ transport format for use on anycomputer with access to SASÒ.

¨ CD-ROM drive: Any PC with a CD-ROMdrive should be able to read the SAF CDs.

¨ Disk storage: Between 10 and 600 mega-bytes are needed for use of the Core CD,depending on the files being used. Thedata on each CD require from 550 to 650megabytes of disk storage. Keep in mindthat you will need space for temporarywork files and for the files you create.

¨ Software: SASÒ. Files converted to SASÒ

transport format can be used by SPSS.¨ People with software experience: The SAF

documentation provides some of thebasics of loading the files into SASÒ andusing them, but further work with the filesrequires SASÒ experience.

CostThe price of the files covers the cost ofreproducing and shipping the file and itsdocumentation, the administrative cost ofhandling the sales, and the cost of technicalsupport to researchers. Checks must be made

payable to the Minneapolis Medical ResearchFoundation. These prices are subject to change.

DocumentationThe Researcher’s Guide to the USRDS Databaseprovides most of the documentation of the SAFs.It includes a codebook of variables on the files,copies of the data collection forms used by theSpecial Studies, and a chapter on techniques forusing the SAFs in SASÒ. Copies of theResearcher’s Guide may be downloaded from theUSRDS website, or requested by phone or e-mail.

ACKNOWLEDGMENT FOR USE OF USRDS DATAPublications that use USRDS data should includean acknowledgment and the following notice:

The data reported here have been supplied by theUnited States Renal Data System (USRDS). Theinterpretation and reporting of these data are theresponsibility of the author(s) and in no wayshould be seen as an official policy or interpreta-tion of the U.S. Government.

DATA RELEASE POLICY & PROCEDURESince the Standard Analysis Files and custom datafiles contain confidential, patient-specific data,release of these files requires the approval pro-cess described here. Investigators may contact theUSRDS Project Officer at the National Instituteof Diabetes and Digestive and Kidney Diseases(NIDDK) to discuss their requests before prepar-ing a written proposal (see the list of USRDS con-tacts in the Forward). To request and use USRDSdata files, investigators should do the following:

¨ Provide the USRDS Project Officer (PO)with a detailed description of the proposedinvestigation (see Table e.4). The projectsummary must include goals, backgrounddata, an in-depth description of the study

Submit to:Lawrence Y.C. Agodoa, MDNIDDKDemocracy 26707 Democracy BlvdBethesda, MD 20892-5458Phone 301.594.7717Fax [email protected]

Paul Eggers, PhDNIDDKRoom 6156707 Democracy BlvdBethesda, MD 20892-5458Phone 301.594.8305Fax [email protected]


design and analytic methodology, and re-sources available for completing theproject, and may be the project descrip-tion from a grant or other funding appli-cation. The proposed project must com-ply with the Privacy Act of 1974, and theproject summary should provide enoughinformation to enable assessment of com-pliance. Guidelines for adherence to thePrivacy Act are contained in the USRDS“Agreement for Release of Data,” providedat the end of this Appendix.

¨ Indicate which USRDS Standard Analy-sis Files will be needed. If the USRDS Stan-dard Analysis Files cannot meet the re-quirements of the proposed research, theproposal must specify precisely which dataelements are needed, and investigatorsmust budget for a substantially higher cost.

¨ If the project is approved, return a signedcopy of the USRDS “Agreement for Re-lease of Data” to the PO. The investigatorand the CC will resolve any technical ques-tions. The investigator will arrange pay-ment with the CC, and payment must bereceived before the files will be released.Checks must be made payable to the Min-neapolis Medical Research Foundation.

The NIH will review the project for technicalmerit and conformity with the Privacy Act. ThePO will notify the investigator(s) in writing ofthe outcome, and if the project is not approvedwill discuss reasons for the decision. The PO willsend a copy of the approval letters to the USRDSCC. The process of reviewing the data request,generating the data file, and releasing the data willtake the CC approximately three months.

When both a copy of the signed “Agreement forRelease of Data” and payment for the files havebeen received by the USRDS CC, the CC will pre-pare the files and documentation and will sendthem to the investigator.

Any reports or articles resulting from use of theUSRDS data must be submitted to the PO priorto submission for publication for review to assureadherence to the Privacy Act. The PO mustrespond within 30 days. If a report or article isdetermined not to adhere to the Privacy Act, itshall not be published until compliance with theAct is achieved. Assessment of compliance will

not depend on the opinions and conclusionsexpressed by the investigators, nor will the PO’sapproval indicate government endorsement ofthe investigator’s opinions and conclusions.

All publications using the released data mustcontain the standard acknowledgement anddisclaimer presented above. The investigator isrequested to send copies of all final publicationsresulting from this research to both the PO andthe USRDS CC.

CaveatsThis policy establishes conditions and proceduresfor the release of data from the USRDS and isintended to ensure that data are made availableto investigators in the pursuit of legitimate bio-medical, cost-effectiveness, or other economicresearch.

The USRDS will not release data that identify in-dividual patients, providers, or facilities. Since itmight be possible, however, to infer the identityof individual patients, providers, or facilities fromthe data in the Standard Analysis Files, the datain these files are considered confidential. TheUSRDS “Agreement for Release of Data” containsa number of both general and specific restrictionson the use of USRDS data, and investigators areexpected to abide by these restrictions. If indi-vidually identifiable data are needed, the requestshould be submitted directly to the Health CareFinancing Administration.

Use of these data to identify and/or contactpatients, facilities, or providers on the files isprohibited both by USRDS policy and by thePrivacy Act of 1974.

The USRDS CC will provide data in any of theusual media (tape, disk, or hard copy). Analyticalservices, other than review of the proposal andpreparation of the data file, will not, however, beprovided under the USRDS contract, althoughUSRDS CC personnel may participate in analy-ses funded by other sources.

Standard Analysis Files or other data files fromUSRDS Special Studies will become available oneyear after the data have been collected, edited, andentered into the database.


United States Renal Data System (USRDS)Agreement for Release of Data

In this agreement, “Recipient” means ___________________________________________________________

________________________________________________________________________________________

A. The National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK), through the United States Renal DataSystem (USRDS) Coordinating Center (CC), will provide the Recipient with tapes, disks, and/or hard copies containingdata extracted from the USRDS research database.

B. The sole purpose of providing the data is the conduct of legitimate and approved biomedical, cost-effectiveness, and/orother economic research by the Recipient.

C. The Recipient shall not use the data to identify individuals on the file.

D. The Recipient shall not combine or link the data provided with any other collection or source of information that maycontain information specific to individuals on the file, except where written authorization has been obtained through theapproval process.

E. The Recipient shall not use the data for purposes that are not related to biomedical research, cost-effectiveness, or othereconomic research. Purposes for which the data may not be used include, but are not limited to,¨ the identification and targeting of under- or over-served health service markets primarily for commercial benefit¨ the obtaining of information about providers or facilities for commercial benefit¨ insurance purposes such as redlining areas deemed to offer bad health insurance risks¨ adverse selection (e.g., identifying patients with high risk diagnoses)

Any use of the data for research not in the original proposal must be approved by the USRDS Project Officer (PO).

F. The Recipient shall not publish or otherwise disclose the data in the file to any person or organization unless the data havebeen aggregated (that is, combined into groupings of data such that the data are no longer specific to any individualswithin each grouping) and no cells (aggregates of data) contain information on fewer than ten individuals or fewer thanfive providers or facilities. The Recipient shall not publish or otherwise disclose data that identify individual providers orfacilities, or from which such identities could be inferred. However, the Recipient may release data to a contractor forpurposes of data processing or storage if (1) the Recipient specified in the research plan submitted to the USRDS ProjectOfficer that data would be released to the particular contractor, or the Recipient has obtained written authorization fromthe PO to release the data to such contractor, and (2) the contractor has signed a data release agreement with the PO.

G. A copy of any aggregation of data intended for publication shall be submitted to the PO for review for compliance with theconfidentiality provisions of this agreement prior to submission for publication and, if not approved, shall not be pub-lished until compliance is achieved. The PO must respond within 30 days.

H. Appropriate administrative, technical, procedural, and physical safeguards shall be established by the Recipient to protectthe confidentiality of the data and to prevent unauthorized access to it. The safeguards shall provide a level of securityoutlined in OMB Circular No. A-130, Appendix III—Security of Federal Automated Information System, which setsforth guidelines for security plans for automated information systems in Federal agencies.

I. No copies or derivatives shall be made of the data in this file except as necessary for the purpose authorized in thisagreement. The Recipient shall keep an accurate written account of all such copies and derivative files, which will befurnished upon request to the PO. At the completion of the activities in the research plan, the file shall be returned to theUSRDS CC at the Recipient's expense, and any derivative files and copies shall be destroyed.

J. Authorized representatives of the PO and/or of HCFA will, upon request, be granted access to premises where data in thisfile are kept for the purpose of inspecting security procedures and arrangements.


___________________________________________________________________________Recipient typed name, title, and organization

___________________________________________________________________________Recipient telephone number

___________________________________________________________________________Recipient signature & date

___________________________________________________________________________Contractor typed name, title, and organization, as appropriate

___________________________________________________________________________Contractor telephone number

___________________________________________________________________________Contractor signature & date

Lawrence Y. C. Agodoa, MD, NIDDK, NIH or

Paul W. Eggers, PhD, NIDDK, NIH

USRDS Project Officer typed name & organization

___________________________________________________________________________USRDS Project Officer signature & date

Revised June 1994

Documents

appendices with eps '01 - USRDS · rity Act of 1965 and in subsequent amendments to the Act. A person in one of these four catego-ries is eligible to apply for Medicare entitlement: