Peritoneal Dialysis International, Vol. 25, pp.

320332
Printed in Canada. All rights reserved.

0896-8608/05 $3.00 + .00

Copyright 2005 International Society for Peritoneal Dialysis

CHARACTERIZING COMORBIDITY IN DIALYSIS PATIENTS: PRINCIPLES OF

MEASUREMENT AND APPLICATIONS IN RISK ADJUSTMENT AND PATIENT CARE

Dana Miskulin
Division of Nephrology, New England Medical Center, Boston, Massachusetts, USA
Comorbid conditions are highly prevalent in dialysis patients and are significant predictors of mortality and other
adverse outcomes. Accordingly, it is important to account
for differences in comorbid illness burden among groups
of dialysis patients being compared. At present, there is
no consensus on what conditions matter, how each should
be defined, and what weights each carries when defining
an individuals risk or case-mix severity. A number of comorbidity instruments, generic or disease specific, have
been employed in dialysis populations. They differ by the
representation and definition of conditions as well as instrument scoring. No instrument has been found to be superior to another in terms of predictive accuracy for
mortality, and accuracy across the board is low. Further
studies are needed to determine whether improvements
would be found with the use of more specifically defined
items and through assignment of item weights based on
relationships for outcomes specifically in a dialysis population. The roles of other factors in risk prediction, such as
markers of nutritional status, inflammation, or other
physiological parameters, relative to comorbid conditions
also need to be defined. Outcomes other than mortality are
likely to identify different factors and/or different relationships than those noted for mortality, which also require
study.
Comorbidity is important for risk adjusting comparative
analyses in nonrandomized trials and quality of care assessments and may, in future, influence payment for dialysis services. Efforts to improve the management of
comorbid illnesses are needed. Comorbid conditions must
be documented accurately and uniformly in all dialysis patients to enable these applications.
Perit Dial Int 2005; 25:320332

www.PDIConnect.com

KEY WORDS: Case-mix severity; case-mix adjustment;

survival analysis; questionnaires; quality of care.
Correspondence to: D. Miskulin, Division of Nephrology,
New England Medical Center, Box 391, 750 Washington Street,
Boston, Massachusetts, 02111 USA.
dmiskulin@tufts-nemc.org
Received 16 February 2005; accepted 5 April 2005.
320

omorbid illnesses, which refer to medical conditions

other than the index disease, that is, kidney failure,
are highly prevalent in the dialysis population, a reflection of the primary disease that caused kidney failure or
the untoward effects of kidney failure on other organ
systems. Cardiovascular disease in particular is common
and is a major cause of the striking increase in mortality
observed across all age groups of patients with chronic
kidney disease (CKD) compared with controls (1,2). The
incident dialysis population in the USA in 2003 illustrates
the tremendous burden of comorbidity of these patients;
over 60% had at least one cardiovascular comorbidity,
with ischemic heart disease in 25%, congestive heart
failure (CHF) in 32%, and peripheral vascular disease in
14% (3). Epidemiological studies have shown that
growth in new end-stage renal disease (ESRD) cases in
the USA in recent years has far outpaced CKD prevalence,
likely reflecting the broader acceptance of individuals
into dialysis programs and the improved prevention and
outcomes of cardiovascular disease, resulting in increased longevity to reach dialysis (46). In keeping with
this, the number and extent of comorbid illnesses in the
average patient initiating dialysis have increased over
the past two decades (710). These observations highlight the need for increased attention to comorbid illnesses, not simply in the context of risk-adjusting clinical
studies, but for a number of reasons, including the dayto-day care of patients.
Comorbidity is a major determinant of morbidity and
mortality, thereby defining an individuals risk for adverse outcomes, that is, case-mix severity (11). The other
case-mix factors involved will depend on the outcome
under study but typically include demographic and socioeconomic factors, physiological parameters, treatment-related factors, and some that may be difficult to
define (e.g., cultural or religious factors) or measure
(e.g., genetic factors). The usual context in which case
mix is an issue is when groups of patients, defined by a
treatment or parameter of interest, are compared for an
outcome but differ inherently by risk factors for that

PDI

JULY 2005 VOL. 25, NO. 4

outcome, not simply by chance but because the same factors led to the treatment or parameter under study in
the first place (Figure 1). This confounding by case mix
must be accounted for in order to determine the true relationship of the treatment or other parameter with the
outcome (12). For example, the factors leading to the
selection of peritoneal dialysis (PD) over hemodialysis
(HD), such as young age, minimal comorbidity, greater
functional ability, and desire for independence in daily
living (1315), also predispose to longer survival. Without adequate control for the propensity to increased survival in PD patients independent of the modality or any
other factor, comparisons of PD versus HD will be biased
in favor of PD.
A large number of potential case-mix factors may differ across groups of dialysis patients, but how much casemix adjustment is enough? In theory, adequacy of
case-mix adjustment is measured as the extent to which
variance in outcome is explained (16); in practice, it is a
function of the number of factors that can be reliably
measured in all subjects. Groups of dialysis patients are
compared frequently in quality-monitoring programs
and clinical studies, but, in most cases, case-mix differences are ignored or inadequately controlled for. Notably, comparisons of mortality rates across dialysis
facilities in the USA account only for differences in age,
race, gender, and the primary diagnosis across populations (17). These factors explain very little variability of
the survival of dialysis patients (18,19). A major hurdle
limiting more routine measurement and adjustment for
case mix is a lack of knowledge about what comorbid conditions matter and how they should be defined and
weighted.
For a host of reasons, not the least of which is the
sheer number of comorbid conditions and combinations
of conditions encountered in a dialysis population that

Figure 1 The relationship of case-mix factors to the treatment and outcome under study.

MEASURING COMORBIDITY IN DIALYSIS PATIENTS

are interrelated, comorbidity is one of the most challenging aspects of case-mix severity to define. This review
will discuss general principles surrounding the measurement of comorbidity in dialysis patients, compare instruments used in past studies, discuss the role of case-mix
factors other than comorbidity in risk adjustment, and
will identify the many applications of comorbidity data
that emphasize the need for uniform comorbidity reporting in all dialysis patients.
SOURCES OF COMORBIDITY DATA

An accurate and up-to-date source of information is

necessary to characterize the presence and extent of an
individuals comorbid illnesses. There are three general
sources of comorbidity data: the patient (self-report),
administrative databases, and review of the medical
record, the latter varying in accuracy depending on the
skill and knowledge of the person abstracting the data.
The advantages and disadvantages of each are outlined.
Self-Report of Comorbid Illnesses: Self-reporting of
medical conditions may be reliable in other populations,
but reservations are warranted, considering the lengthy
and complex medical histories typical of dialysis patients as well as the high prevalence of cognitive deficiencies, which are estimated to affect more than 25%
of patients (20,21). A single study compared self-reporting of 8 comorbid conditions against data obtained
through a systematic medical record review in 959 incident dialysis subjects at baseline in the CHOICE
(Choices for Healthy Outcomes in Caring for ESRD) Study
(22). Specificity was high, meaning patients did not
report diseases that they did not have; however, sensitivity was low, even though these were major conditions
such as CHF, myocardial infarction, stroke, and chronic
obstructive pulmonary disease. Self-report of symptoms
or limitations from disease, as opposed to diagnoses,
are collected with health status assessments and may
be more promising, but also have drawbacks, as will be
discussed later.
Administrative Databases: Administrative databases
serve as a source of comorbidity information for large
populations; the major advantage being that the data
already exist, reducing time and expense that would be
otherwise spent in chart review. Diagnoses and procedures stemming from hospitalizations and other encounters are coded using uniform taxonomies such as the
International Classifications of Disease, 9th revision.
Some drawbacks relate to the coding systems themselves
in that they are not specific for a dialysis population,
and where an exact code is not present, the closest substitute may be inappropriate or too broad in definition
321

MISKULIN

to accurately depict the true diagnosis. Other inaccuracies may arise because personnel coding the medical
records are not medically trained and codes with higher
reimbursements may be preferentially used. National
registries, such as the United States Renal Data System,
are another form of administrative database, now
present in several countries, that systematically collect
data to study epidemiology and outcomes of the ESRD
population over time. In the USA, no quality control procedures govern the collection of comorbidity, undertaken in new Medicare-entitled patients using the
Medical Evidence Form (Form 2728), which captures the
presence or absence of 18 major conditions and 2 physical impairments (23). Longenecker et al. compared data
from the Form 2728 against chart reviews in subjects
enrolled to the CHOICE Study and revealed gross
underreporting, even for major comorbid conditions
(24). Efforts to improve the collection and quality of comorbidity data within national data systems should be
strongly encouraged, as these data will enable standardized measurements and comparisons within and between
national populations.
Medical Record Review: The medical record is the means
by which health care providers communicate with one
another and is generally considered the most reliable
source of information about comorbid conditions (11).
The latter is true if a complete record can be assembled,
which is not a trivial task considering the potential for
records to be dispersed among the offices of the nephrologist, primary care physician, and various consultants,
as well as one or more hospitals and the dialysis unit.
Another source of variability stems from the reviewers
familiarity with the diagnoses and procedures encountered in dialysis patients. In theory, the nephrologist is
the ideal person to compile and maintain the list of past
and active medical problems, although this has not been
the practice in the USA, perhaps because it is time-consuming and benefits may not be readily apparent. Maintaining accurate data requires some investment of time
but is lessened if problem lists are updated as events
occur.
INSTRUMENT CONTENT: REPRESENTATION AND DEFINITIONS
OF COMORBID CONDITIONS

A large number of comorbid conditions and/or combinations of conditions are possible in a dialysis population, but only a finite list can be practicably and reliably
collected on a routine basis. A standardized instrument
is required to ensure consistency. Comorbidity instruments vary considerably in their representation and definition of conditions. To enhance consistency in
322

JULY 2005 VOL. 25, NO. 4

PDI

interpretation across multiple reviewers, conditions

should be defined as objectively as possible, using criteria appropriate for a dialysis population. The detail by
which items should be defined is unclear. Given the high
prevalence of many of the conditions in a dialysis population, it would seem especially important to delineate
subjects at one versus the other end of the severity spectrum for each condition, as prognosis would be expected
to differ greatly between them. As shown in Table 1, when
the broad entity any history of CHF is compared with
the more specific definitions collected in a large HD
population using the Index of Coexistent Diseases (ICED)
(27), the 1-year mortality risk varies considerably, depending on the definition used. A substantial loss of information and of discriminatory power would result if
we considered only the term any CHF in defining an
individuals mortality risk. For risk measurement purposes, defining the severity of each condition however,
would be unnecessary if those with a severe disease (e.g.,
severe CHF) also consistently are afflicted with other
conditions (e.g., ischemic heart disease, diabetes, peripheral vascular disease), in which case, the increase
in risk is captured by the presence of the other conditions. The latter remains to be proven.
A final principle to consider when selecting a comorbidity instrument is the availability of the data used to
define each condition. Non-routine test results in a dialysis unit, such as electrocardiograms or chest x rays,
take time to locate or may be missed. When reviewed retrospectively, the circumstances surrounding the test may
not be apparent and may not accurately reflect the
chronic activity and prognosis of the condition (e.g., an
echocardiogram immediately after an acute myocardial
infarction). Although the mere presence or absence of a
test might in itself be considered a measure of severity,
it may simply reflect differences in physician practice
style, test availability, or that it is missing from the medical record, all of which have nothing to do with the presence or severity of disease (11).
ITEM WEIGHTS AND SCALING

There are two general ways to adjust for case-mix factors. If there is a large number of outcomes, the comorbidity and other case-mix factors can be treated as
separate covariates in the analysis, the advantage being
that the relationships of each are estimated for this study
population as opposed to using predetermined weights
of a comorbidity instrument. More often, it is desirable
to conserve statistical power for assessing the main factor of interest with the outcome so information is summarized into a single covariate using a comorbidity

PDI

JULY 2005 VOL. 25, NO. 4

MEASURING COMORBIDITY IN DIALYSIS PATIENTS

TABLE 1
Relationships of Congestive Heart Failure (CHF) with One-Year Mortality Vary Considerably, Depending on Definition
Definition of CHF
Any history of CHF pre or post ESRD
Cardiomegaly on chest x ray or left ventricular hypertrophy by ECG or ECHO
1 hospitalization in the past year for CHF
2 hospitalizations in the past year for CHF
Frequent intradialytic hypotension
Functional impairments due to dyspnea
Mild-to-moderate
Severe
Predialysis systolic blood pressure
<120 mmHg (reference >140 mmHg)
120140 mmHg

Prevalence (%)

Hazard ratio (95% CI)

45
45
12
1.2
0.6

1.38 (1.191.61)
1.12 (0.961.30)
1.96 (1.622.37)
2.02 (1.193.43)
1.77 (0.843.73)

19
2

1.78 (1.621.96)
3.01 (2.553.56)

8
23

2.61 (2.093.25)
1.60 (1.351.90)

CI = confidence interval.
Unadjusted relationships based on Cox proportional hazards regression analyses conducted in a sample of 3863 patients from
3 hemodialysis populations (12,25,26). Discriminatory power would be reduced if any history of CHF is used without further
defining severity, as denoted in the subsequent items shown. The conditions are correlated with one another and with other
comorbidities. Those that would remain in an adjusted model would depend on the other factors included.

instrument. Depending on the instrument, the comorbidity data may be summarized as the count of conditions per patient or the sum of weighted items. Item
weights are based on relationships with a specific outcome of interest, preferably determined in the population of interest. Weights effectively represent the
prognostic significance of one condition relative to another. The summary score is usually divided into levels
to provide a simple and practical means for risk stratifying a population. The principles regarding instrument
content and scoring are illustrated in comparing four
comorbidity instruments, two of which are generic, the
Charlson Comorbidity Index (CCI) and the ICED, and two
designed specifically for dialysis patients, the Wright
Khan Index and the Davies Index.
COMORBIDITY INSTRUMENTS IN USE IN DIALYSIS
POPULATIONS

Charlson Comorbidity Index: The major appeal of this

instrument is its simplicity. The instrument consists of
19 items, each weighted for mortality as determined from
its original development in a general medical inpatient
population in a New York hospital in the late 1980s, later
validated in a population with breast cancer (28). Item
weights vary from 1 to 6 and the weighted items are
summed to a scale that ranges from 2 to 37 plus the addition of 1 point for each decade after age 40 years. The
time spent scoring a CCI through data abstraction from
chart review in an incident PD population ranged from
10 to 20 minutes in one study (29).

A number of studies have shown increasing CCI score

correlates with increased mortality (2931) and costs
(32). For example, the average CCI score in 268 incident
PD patients from a single USA center at the start of dialysis was 5.4 2.2, and for each increase of 1 in the CCI
score, there was a 50% increase in the subsequent relative risk (RR) of death (RR 1.54; 95% confidence interval 1.36 1.74) (29).
It is, however, reasonable to question whether an instrument designed for a general medical population is
applicable to a dialysis population. First, the medical
conditions found to be most prognostic for mortality
among inpatients on a medical ward in the 1980s are not
likely to be the same as for a present-day outpatient dialysis population. Notably, there are only two items for
cardiac disease, each of which receive the lowest weight,
whereas 6 of 18 items (excluding moderate-to-severe
renal disease) and more than 50% of the score relate to
conditions that, together, account for <5% on average
of an outpatient dialysis population, consisting of metastatic cancer, lymphoma, leukemia, severe liver disease,
AIDS, and hemiplegia. Second, the prognostic significance of conditions as reflected in item weights are likely
to be different in a dialysis population from a general
medical population.
To determine whether CCI items should be weighted
differently, a recent study assessed the relationships of
CCI items with mortality in 237 incident dialysis patients
with an average CCI score of 4.70 2.15 (SD) (30). Several CCI items were found to have different associations
with mortality, as reflected in adjusted item weights in
323

MISKULIN

JULY 2005 VOL. 25, NO. 4

Table 2. As expected, cardiovascular conditions are given

more prominence in the dialysis population. An interesting point, the condition diabetes with complications received a new weight of 1, versus 2 in the original
CCI. Diabetes with complications is common in the dialysis population and, defined as such, this condition
alone is unlikely to effectively discriminate with respect
to survival. Also, the increased risk is captured in its complications peripheral vascular disease, ischemic heart
disease, etc. Oddly, diabetes without complications was
given a weight of 2 over 1, which the authors attributed
to older age of these patients, although there may have
been a problem with sample size as confidence intervals
were wide. Even using the reassigned weights, predictive accuracy was on the low side, leading us to question
TABLE 2
Comparison of Original Charlson Comorbidity Index Weights
and New End-Stage Renal Disease (ESRD) Comorbidity Index
Weights Based on Cox Proportional Hazards Model

Comorbidity variable
Myocardial infarction
Congestive heart failure
Peripheral vascular disease
Cerebral vascular disease
Dementia
Chronic lung disease
Rheumatologic
Peptic ulcer disease
Mild liver disease
Diabetes without complications
Hemiplegia
Neoplasia
Moderate/severe liver disease
Metastatic disease
Leukemia
Lymphoma
Human immunodeficiency virus
Renal disease

Original
weight
1
1
1
1
1
1
1
1
1
1
2
2
3
6
2
2
6
2

New ESRD
comorbidity
weight
2
2
1
2
1
1
1
1
a

2
a

NS
2
10
2
5
a

NI

NS = not significant; NI = not included.

a Insufficient number of patients.
The relationship of individual conditions of the Charlson
Comorbidity Index with mortality was determined in 237 incident hemodialysis patients. Many of the weights changed from
their designation based on relationships in general medical
patients.
Source: Hemmelgarn BR, et al. Adapting the Charlson Comorbidity Index for use in patients with ESRD. Am J Kidney Dis 2003;
42(1): p. 128. Copyright (2003), with permission from the
National Kidney Foundation.
324

PDI

whether the problem lies in the need for more specific

definitions than those on the CCI. One distinct advantage of the CCI is that its widespread use enables direct
comparisons across different populations (3335). The
small number of broadly defined items required to score
this instrument, however, can easily be collected in addition to those required of another instrument if another
proves superior.
Davies Index: This instrument was developed from 97 PD
patients followed for 30 months and consists of 7 equally
weighted items: ischemic heart disease, peripheral vascular disease (including cerebrovascular), malignancy,
left ventricular dysfunction, diabetes, systemic collagen
vascular diseases, and other conditions. The basis for
assigning equal weights to each condition is unclear, but
the relationships with mortality were not equivalent in
its development (36). Furthermore, definitions in some
cases are vague or subjective, for example left ventricular dysfunction as clinical evidence of CHF not attributable to errors in fluid balance, requires clinical judgment
and does not define severity of the condition or frequency
of events, all of which leads to variability in scoring across
multiple reviewers. Nonetheless, several studies, including a multicenter study, the Netherlands Cooperative
Study on the Adequacy of Dialysis (NECOSAD), consistently
show a graded mortality risk for increasing levels of the
score divided into three levels, which persisted even after
adjustment for a number of demographic, physiological,
and treatment-related factors (3638).
WrightKhan Index: The Wright-Khan Index was developed in 375 dialysis patients followed at least 2 years and
divides the population into three risk levels based on age
and the presence of comorbid conditions. These consist
of (1) age <70 years with no comorbidity; (2) age 70 80
with no comorbidity, or age <70 with diabetes or any age
with one comorbidity; and (3) age >80, or any age with at
least two comorbidities, or any age with cardiopulmonary
disease or visceral cancer (39). The lack of specificity
about which comorbid conditions should be counted, that
comorbid conditions and chronological age are considered to have the same prognostic significance, and the
minimal requirements to reach level 3, considering the
high prevalence of comorbid conditions in the dialysis
population, would seem to limit the distribution of patients across score levels and the discrimination among
them. Nonetheless, it too has been deemed a valid mortality predictor because increased risk levels correlated
with mortality in past studies (18,25,39).
Index of Coexistent Disease: The instruments discussed
up to now consist of a limited number of conditions, defined broadly and weighted equally with the exception
of a few items on the CCI. The characterization of co-

PDI

JULY 2005 VOL. 25, NO. 4

morbid illnesses with the ICED is much more extensive,

consisting of over 160 items that denote the presence
and severity of each condition. The instrument has two
parts: the first is based on data abstracted through chart
review and consists of 19 disease categories, defined
further for three grades of disease severity; the second
part of the instrument is an observer-based assessment
of physical impairments. The instrument was originally
designed for a general medical population to predict
functional status outcomes (40), but enhancements,
particularly to the cardiovascular categories, were made
prior to its use in the HEMO Study (19).
The ICED has been used in a number of multicenter settings, including the Hemodialysis (HEMO) Study (41), the
CHOICE study (14), and a not-for-profit national dialysis
provider in the USA (42). The instrument divides the population into three levels and each corresponds with increasing mortality risk with little overlap among levels
(19,26,43). It is the only instrument, to my knowledge,
in which reliability of scoring has been tested, as carried
out in the HEMO Study where the average agreement in
assigning an ICED score to 6 charts across nurses at the
15 study sites and a physician (considered the gold standard) was >90% (41). In addition, it is sensitive to change.
Annual ICED measurements were performed as part of the
CHOICE Study and the change or absence of change in ICED
score was a strong predictor of mortality, independent of
the baseline score (43). Another unique feature of the
ICED is the inclusion of physical impairments, which have
consistently been shown to be strong predictors of mortality and adverse outcomes (23,4446).
The obvious tradeoff for the amount of data collected
is the time taken per assessment, which averages 50 minutes per patient and makes it impractical for everyday use.
Other drawbacks are that the wealth of information collected is lost by summary into only three levels, the prognostic significance of each condition is effectively
considered equivalent because the final score is based on
the single peak disease score combined with the single
peak impairment score, and there is no increment in score
for patients with multiple severe comorbid conditions.
WHICH INSTRUMENT IS BEST AT PREDICTING MORTALITY?

All of the instruments above work in the sense that

incremental risk scores correlate with increased mortality; however, chronological age divided into deciles would
produce a similar result but is a poor mortality predictor
on its own. Measures of model fit (e.g., chi-square test)
statistics are often reported and describe how well the
covariates explain the variability in the outcome but are
difficult to conceptualize in clinical terms. The area under

MEASURING COMORBIDITY IN DIALYSIS PATIENTS

the receiver-operator-characteristic (ROC) curve provides

a measure of discriminatory ability to compare instruments (47). Values for ROC range from 0 to 1 and may be
interpreted as the frequency with which the instrument
assigns a higher score to the person who dies versus survives, given all possible pairs in which one subject dies
and survives. An instrument with an ROC of 1 perfectly
discriminates among those who die versus survive.
van Manen et al. compared the CCI, the Khan, and the
Davies indices for their accuracy in predicting mortality
in 1205 incident PD patients with a 1-year survival rate
of 87% from NECOSAD (31). In addition to the data collected to score these instruments, a separate aim of this
study was to assess whether consideration of disease
severity for four conditions (diabetes, cancer, angina,
and CHF) improved accuracy over the broad categorizations of diseases from the above instruments. Results of
concordance statistics (analogous to ROC) for mortality
predictions of each comorbidity instrument plus age were
roughly equivalent as follows: 0.72 for the Khan, 0.73
for the Davies, 0.74 for the CCI, and 0.75 for the new
index that accounts for disease severity. The addition of
explicit severity definitions for the comorbid conditions
did not appear to improve predictive accuracy, although
a limited number of conditions were defined more
specifically.
In another study, involving 2388 individuals from a
national dialysis provider in the USA, with an annual survival rate of 79%, the CCI, ICED, Khan, and Davies indices were compared for 1-year mortality predictions (48).
A comparison of survival models composed of each comorbidity measure plus age is shown in Table 3. In terms
of discriminatory ability, the ICED appeared superior to
the others when predictions were based simply on a comorbidity instrument plus age. The addition of race, gender, cause of ESRD, vintage, and serum albumin to each
model improved discrimination for each model, and the
margin in favor of the ICED lessened. An ROC of 0.77 is
considered reasonably good for a predictive tool, but
still, 23% of the time, individuals are misclassified as
being at higher risk than an individual who dies, when
they in fact survive, or vice versa. The most widely used
prognostication system, the APACHE, usually yields an
ROC of 0.80 or higher for 30-day mortality (49,50); however, this is probably not a fair benchmark as accuracy
would be expected to be better with a shorter time span
between the measurement and the outcome. Another
important measure of accuracy is calibration, which summarizes agreement between predictions and observed
outcomes. Calibration is particularly important for direct applications of the predictions, for example in reporting an individuals predicted mortality to aid
325

MISKULIN

JULY 2005 VOL. 25, NO. 4

PDI

TABLE 3
Predicting One-Year Mortality: a Comparison of Comorbidity Indices + Age

Instrument (+ age)a
ICED
01
2
3
CCI
1
2
3
4
5
WrightKhan
1
2
3
Davies
1
2
3

Subjects within
each level [n (%)]

1-year
mortality (%)

545 (31)
500 (28)
734 (41)

9.7
23.2
36.1

181 (10)
116 (7)
767 (43)
507 (29)
208 (12)

6.1
10.3
21.2
30.3
45.2

204 (12)
787 (33)
992 (56)

4.9
18.9
31.7

301 (17)
937 (53)
541 (30)

9.6
22.8
35.3

Unadjusted
Area under
ROCb curve
95% CI

Adjustedc
Area under
ROCb curve
95% CI

0.72

0.690.75

0.77

0.750.79

0.67

0.650.70

0.74

0.720.77

0.68

0.650.70

0.75

0.720.78

0.68

0.650.70

0.75

0.730.78

ICED = Index of Coexistent Disease; CCI = Charlson Comorbidity Index; ROC = receiver operator characteristic; CI = confidence
interval.
a Age was added to each model.
b The ROC provides a measure of discriminatory ability and was highest for the ICED when a comorbidity measure was used on its
own to predict mortality. Upon the addition of other case-mix factors, the discriminatory ability of each model was equivalent,
regardless of the comorbidity measure.
c Each model was adjusted for age, comorbidity measure, race, gender, cause of end-stage renal disease, and serum albumin.
The proportion that died within a year increased with higher risk score for each comorbidity instrument. [Adapted from Ref. (48)]
Source: Miskulin D, et al. Predicting one-year mortality in an outpatient hemodialysis population: a comparison of comorbidity
instruments. Nephrol Dial Transplant 2004; 19(2): p. 416. Copyright (2004). Reprinted with permission from the European Renal
AssociationEuropean Dialysis and Transplant Association.

decision-making. Results show the WrightKhan Index

may not be calibrated, but the others were satisfactory.
The list of instruments above is by no means a complete list of comorbidity measures used in dialysis patients, although, to my knowledge, the accuracy of the
others has not been tested. We can conclude that all of
these instruments are roughly equivalent once other
case-mix factors are added, but none is sufficiently accurate to be used alone in making clinical decisions.
ALTERNATE APPROACHES TO COMORBIDITY AND/OR CASE-MIX
MEASUREMENT

Physical Impairments and Health Status Measurements:

In clinical practice, we gauge the severity of many conditions by the patients accounts of the intensity or frequency of symptoms or their impact on functionality.
326

Self-report of disease severity through symptom burden

is the basis of instruments successfully used in general
medical, diabetic, and ischemic heart disease outpatient
populations (51,52). In the dialysis population, a number of studies have shown strong relationships of health
status or physical impairments with mortality, including through self-report with the SF-36 (53,54), or as reported by observers, for example with the Karnofsky
Index (44,46), the ICED (43), or Form 2728 (23). While
these approaches intuitively make sense and would be
less costly than collection of comorbid conditions, major
consideration should be given to the fact that the individuals at greatest risk (and those most important to
capture), may be selectively omitted because they are
too ill and/or cognitively impaired to respond. Observerbased assessments should also be viewed with caution
given the subjective nature of scoring, which may pose

PDI

JULY 2005 VOL. 25, NO. 4

problems depending on the application. If, for example,

resource allocation was tied to case mix or through risk
adjustment, facility performance reports could be made
to look better, there would be incentive to overstate the
illness burden of ones dialysis population.
FACTORS OTHER THAN COMORBIDITY AS RISK ADJUSTORS

Unique to dialysis populations over other outpatient

populations is the amount of clinical data collected on a
routine basis. A number of studies have demonstrated
relationships of various laboratory and physiological parameters, for example serum albumin, systolic blood
pressure, body mass index, serum phosphate, and parathyroid hormone, with mortality and other outcomes
in dialysis patients (23,5558). Markers of malnutrition
and inflammation, such as interleukin-6 (IL-6), C-reactive protein (CRP), serum ferritin, total cholesterol, and
others also associate strongly with mortality (59,60).
Although it may be unclear why a patient has a high CRP
or a low serum albumin, from a risk stratification standpoint, that the factor is highly predictive for mortality
and is reliably measured is all that matters. Many of these
factors are already collected as part of the monthly or
quarterly blood work in HD and PD populations. Given
the expense of collecting comorbidity information and
the potential for variability, at any point from compiling
the data record to scoring it, laboratory tests are more
and more appealing. For some of these laboratory measures, factors affecting the reliability of measurement,
such as diurnal variation, timing relative to dialysis, or
upsets due to an acute illness in an otherwise low-risk
individual, need to be worked out before they can be proposed as risk stratification measures for widespread use
(61). Further, whether comorbidity can be substituted,
in part or wholly, by inflammatory and physiological parameters for risk stratification purposes remains to be
tested. One study showed this was not the case in that
comorbidity, as measured using the ICED, maintained
significance and contributed more to the survival model
than either CRP or IL-6, or the combination (62), although further studies to address this are needed. One
thing is for certain, these nonspecific inflammatory and
nutritional markers cannot replace the need to characterize comorbid conditions for the purposes of patient
care.
STRATEGIES FOR IMPROVING PREDICTIVE ACCURACY

In reviewing studies relating a comorbidity instrument

and other factors with mortality, four issues need to be
explored as a means to improve risk predictions in dialy-

MEASURING COMORBIDITY IN DIALYSIS PATIENTS

sis patients. First, the comorbid conditions in the above

instruments are defined broadly, or, as in the case of the
ICED, the information regarding disease severity is effectively lost in the final scoring. The use of explicit definitions for disease severity for each condition is likely to
improve discrimination. Second, the weights of conditions in the CCI are based on relationships with mortality in a general medical population, and in the other
instruments, purely on clinical judgment. Relationships
of factors with mortality must be worked out specifically
in the dialysis population. Third, the contribution of
laboratory and physiological data is strong and much of
the data is routinely available in a dialysis population.
Such factors should be considered alongside the comorbid conditions to identify and weight those of key prognostic significance as the physiological data and
comorbidity data are likely to be correlated. Fourth, comorbidity and other selected factors will change with
time, yet these changes cannot be incorporated in a predictive instrument because it is not known with certainty
who will survive to the next interval at which variables
are updated. To account for this, the maximum time span
(between assessment and outcome) that ensures stability and accuracy of predictions needs to be worked out.
It may be less than a year, but cannot be so short that it
would be impractical to update the variables at the required frequency in routine practice. To conduct these
studies, a large population with an extensive list of candidate comorbid conditions and other case-mix factors
is needed.
USES OF COMORBIDITY DATA

Risk Adjustment in Clinical Research: Case-mix severity

is generally considered a nuisance factor for nonrandomized studies; there is no particular interest in the
case-mix factors themselves but, in order for the treatment effect to be measured accurately, differences in
study arms must be controlled for. Even in randomized
trials when small differences that are strongly prognostic for the outcome exist, adjustment for case mix can increase statistical power to detect the treatment effect
(63,64). The potential for error from lack of control over
case-mix bias is illustrated in a recent study of the effects
of cholesterol on mortality (65). It has long been questioned whether measures to lower cholesterol should be
applied to the dialysis population, given past observations of an increase in mortality risk with lower cholesterol and a decrease with higher cholesterol (66,67). The
term reverse epidemiology has been coined to describe
other instances where relationships of nutritional or cardiovascular risk factors with mortality are opposite to
327

MISKULIN

those seen in the general population (6870). A recent

study showed that, once the population is stratified by
case mix (into high and low groups), effectively controlling for the fact that those with lower cholesterol were
inherently sicker, a different result emerged. In those at
low-risk, a decreasing cholesterol was associated with
reduced mortality, consistent with relationships in the
general population. A similar effect of modification by
case mix was found with body mass index and mortality
(71). Reverse epidemiology may be nothing more than inadequate control over case-mix severity. Increased numbers of clinical trials are being conducted in CKD
populations and greater attention to case mix is needed
to avoid erroneous conclusions.
Comorbid illnesses may lead to different responses in
treatment. For example, PD may be postulated to confer
a survival advantage for patients with severe left ventricular dysfunction because of the continuous nature
of ultrafiltration and avoidance of hemodynamic insults
of HD. When examining a treatment effect, these patients need to be separated from the rest of the PD population as they may respond very differently. Recent
studies using administrative USA data show specific subgroups, including new ESRD patients with CHF, ischemic
heart disease, or large body habitus, have, if anything,
reduced survival compared with their counterparts on
HD and in contrast with the rest of the population
(13,72,73). Further studies are needed to determine
whether these findings are reproduced and, if so,
whether different treatment targets or other modifications are needed in the setting of specific comorbid
illnesses.
Risk Adjustment in Quality Improvement Programs: The
Clinical Performance Measures Program in conjunction
with the National Kidney Foundation sets and monitors
adherence to various parameters of care and, in doing
so, strives to improve the quality of care of patients receiving dialysis at facilities throughout the USA (74).
Case-mix adjustment may be integral to quality reporting given that there are well-known differences in patient populations across providers that may affect
process and outcome measures (52,75). In a facility with
higher than average case mix, it is likely to be more difficult to adhere to the designated quality benchmarks,
resource use is higher, and survival rates lower. Without
accounting for such differences, irrespective of the care
delivered, units caring for sicker patients will be scrutinized unfairly, while those caring for healthier populations pass unnoticed (76). Providers are increasingly held
accountable for their results by patients and payers. With
increased pressure to achieve results, and with ever dwindling resources, dialysis facilities may be pressured into
328

JULY 2005 VOL. 25, NO. 4

PDI

cherry picking the younger and healthier patients,

which is far from the goals of a quality improvement program (77). Control over case-mix differences is critical
to the assurance that the measures being monitored actually represent quality (as opposed to inherent differences in patient populations) and thereby promise
improved care and outcomes.
Resource Allocation: Whether it be warfarin dosing,
antibiotic administration, diabetes management, relaying care plans to caretakers in institutionalized patients,
or the extra dialysis sessions required of patients prone
to intradialytic hypotension, there are ample reasons
why patients with greater burden of comorbid illness cost
more to care for. This, unfortunately, is not reflected in
the current reimbursement for outpatient dialysis services in the USA. In order for a case-mix adjusted model
for reimbursement to become operational, the key comorbid conditions that are influential will need to be
identified and measured in all dialysis patients.
Management of Comorbid Conditions in Everyday Practice: Studies have shown that fewer patients with CKD than
in the general population receive treatments considered
the standard of care for cardiovascular disease, cardiovascular risk factors, diabetes and lipid management, and
other preventive measures (3,78,79). Accordingly, the
National Kidney Foundation and others are expanding
clinical practice guidelines to the management of diabetes and lipid and cardiovascular disease in CKD patients
(80). Disseminating these into practice poses a greater
challenge. As a first step, documentation of past and current medical conditions in the dialysis unit medical record
needs to improve (Figure 2). This alone will enhance communications among the multidisciplinary team of providers and will lead to better treatment decisions. The use of
standardized care algorithms for specific comorbid conditions encountered in a dialysis population is a novel
approach to improving comorbid illness management in
a dialysis population (81). A recent study, the Dialysis Risk
Factor Intervention Trial, developed and tested the use
of standardized algorithms for management of blood
pressure, diabetes, lipids, ischemic heart disease, atrial
fibrillation, and CHF in five dialysis units of a not-forprofit provider in the USA (82,83). The algorithms focused
on pharmacological and nonpharmacological treatments
and were based on review of the medical evidence and,
where they existed, clinical practice guidelines. A pharmacist reviewed a standardized comorbidity profile and
the medication records for each patient, and recommendations were communicated to the physician, who could
decide to accept or reject them on a case-by-case basis.
Recommendations were followed 80% of the time and
were found to improve care (defined as the addition of a

PDI

JULY 2005 VOL. 25, NO. 4

MEASURING COMORBIDITY IN DIALYSIS PATIENTS

Figure 2 Collection and uses of comorbidity data in practice.

standard treatment without adverse effects), have no

impact, or worsen care 89.9%, 7.6%, and 2.4% of the time,
respectively. Whether this program impacts on hard clinical end points requires further follow-up. An electronic
medical record system would enable automation of care
algorithms for implementation and monitoring in large
populations.
In order for a population-based strategy such as the
above to be implemented, individuals with a specific condition or risk factor must be accurately identified and
not missed, another reason why a standardized assessment of comorbidity should be routine. The list of comorbid conditions identif ied as key predictors of
mortality and used for risk-adjustment purposes will be
much smaller and may be defined differently (more specifically, probably) than those addressed by the care algorithms. A standardized assessment of comorbidity
should include the conditions as defined appropriately
for both applications.
SUMMARY
There are many reasons why comorbid conditions
should be measured in a standardized fashion in all dialysis patients: to enable unbiased comparisons in clinical tr ials and quality improvement programs, to
implement strategies for improving comorbid illness
management, to provide projections about prognosis
that aid decision-making, to enable more equitable allocation of resources in order that the current standards
of care can continue to be achieved. A simple yet accurate instrument that collects the pertinent information
for risk stratification purposes and disease management
initiatives is required. Studies are needed to identify and
weight the key conditions that relate to mortality, also
after consideration of other demographic and physiological variables that are strong predictors and would

be routinely available in a dialysis population. Different

factors and weights may apply for different outcomes and
also deserve study. In the interim, local and national
strategies to improve the documentation and management of comorbid conditions are greatly needed, as such
efforts may yield improved care and outcomes.
REFERENCES
1. Sarnak M, Levey A. Cardiovascular disease and chronic
renal disease: a new paradigm. Am J Kidney Dis 2000; 35(4
Suppl 11):S11731.
2. Foley R, Parfrey PS, Sarnak MJ. The clinical epidemiology
of cardiovascular disease in chronic renal disease. Am J
Kidney Dis 1998; 32(5 Suppl 3):S11219.
3. United States Renal Data System. Excerpts from the USRDS
2004 annual renal data report. Am J Kidney Dis 2005;
45(1 Pt 2):1280.
4. Xue J, Ma J, Louis T, Collins A. Forecast of the number of
patients with end-stage renal disease in the United States
to the year 2010. J Am Soc Nephrol 2001; 12:27538.
5. Muntner P, Coresh J, Powe N, Klag M. The contribution of
increased diabetes prevalence and improved myocardial
infarction and stroke survival to the increase in treated
end-stage renal disease. J Am Soc Nephrol 2003; 14:
156877.
6. Hsu C-Y, Vittinghoff E, Lin F, Shlipak MG. The incidence of
end-stage renal disease is increasing faster than the
prevalence of chronic renal insufficiency. Ann Intern Med
2004; 141:95101.
7. Port F. The end-stage renal disease program: trends over
the past 18 years. Am J Kidney Dis 1992; 20:37.
8. Mailloux LU, Napolitano B, Alessandro G, Bellucci G,
Mossey RT, Vernace MA, et al. The impact of co-morbid risk
factors at the start of dialysis upon the survival of ESRD
patients. ASAIO J 1996; 42:1649.
9. Hylander B, Lundblad H, Kjellstrand C. Changing patient
characteristics in chronic hemodialysis. Scand J Urol
Nephrol 1991; 25:5963.
329

MISKULIN

10. Collins AJ, Hanson G, Umen A, Kjellstrand C, Keshaviah P.

Changing risk factor demographics in ESRD patients entering hemodialysis and the impact on long-term mortality. Am J Kidney Dis 1990; 15:42232.
11. Iezzoni L. Risk Adjustment for Measuring Health Outcomes. 2nd ed. Ann Arbor, MI: Health Administration
Press; 1997.
12. Greenfield S, Aronow HU, Elashoff RM, Watanabe D. Flaws
in mortality data: the hazards of ignoring comorbid disease. JAMA 1988; 260:22535.
13. Stack A. Determinants of modality selection among incident US dialysis patients: results from a national study.
J Am Soc Nephrol 2002; 13:127987.
14. Miskulin DC, Meyer KB, Athienites NV, Martin AA, Marsh
JV, Fink NF, et al. Comorbidity and other factors associated with modality selection in incident dialysis patients:
the CHOICE Study. Am J Kidney Dis 2002; 39:32436.
15. Collins A, Weinhandl E, Snyder J, Shu-Cheng C, Gilberston
D. Comparison and survival of hemodialysis and peritoneal dialysis in the elderly. Semin Dial 2002; 15:98102.
16. Greenfield S, Sullivan L, Silliman RA, Dukes K, Kaplan SH.
Principles and practice of case-mix adjustment: applications to end-stage renal disease. Am J Kidney Dis 1994;
24:298307.
17. Wolfe R, Gaylin D, Port F, Held P, Wood C. Using USRDS
generated mortality tables to compare local ESRD mortality rates to national rates. Kidney Int 1992; 42:9916.
18. Khan I, Campbell M, Cantarovich D, Catto G, Delcroix D,
Edward N, et al. Comparing outcomes in renal replacement
therapy: how should we correct for case mix? [Published
erratum appears in Am J Kidney Dis 1998; 31:900]. Am J
Kidney Dis 1998; 31:4738.
19. Athienites NV, Miskulin DC, Fernandez GF, Bunnapradist
S, Simon G, Landa M, et al. Comorbidity assessment in
hemodialysis and peritoneal dialysis using the Index of
Coexistent Disease (ICED). Semin Dial 2000; 13:3206.
20. Sehgal A, Grey S, DeOreo P, Whitehouse P. Prevalence, recognition, and implications of mental impairment among
hemodialysis patients. Am J Kidney Dis 1997; 30:419.
21. Kurella M, Chertow G, Luan J, Yaffe K. Cognitive impairment in chronic kidney disease. J Am Geriatr Soc 2004;
52:18639.
22. Merkin SS, Longenecker JC, Fink NE, Levey AS, Powe NR.
Comparison of patients reports of comorbid disease with
the medical record [Abstract]. J Am Soc Nephrol 2000; 11:
157A.
23. United States Renal Data System. USRDS 1991 annual data
report: comorbid conditions and correlations with mortality risk among 3,399 incident hemodialysis patients.
Am J Kidney Dis 1992; 20:328.
24. Longenecker JC, Klag MJ, Coresh J, Levey AS, Martin AA,
Fink NE, et al. Validation of comorbid conditions on the
end-stage renal disease medical evidence report: the
CHOICE Study. J Am Soc Nephrol 2000; 11:5209.
25. Khan I, Graeme R, Edward N, Fleming L, Henderson I,
MacLeod A. Influence of coexisting disease on survival on
330

JULY 2005 VOL. 25, NO. 4

PDI

renal-replacement therapy. Lancet 1993; 341:41518.

26. Nicolucci A, Cubasso D, Labbrozzi D, Mari E, Imicciatore P,
Procaccini DA, et al. Effect of coexistent diseases on survival of patients undergoing dialysis. ASAIO J 1995; 41:
M291M295.
27. Miskulin D, Martin A, Brown R, Levey A. Development and
validation of a dialysis-specific comorbidity index [Abstract]. J Am Soc Nephrol 2002; 13:613A.
28. Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new
method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis
1987; 40:37383.
29. Fried L, Bernardini J, Piraino B. Charlson Comorbidity
Index as a predictor of outcomes in incident peritoneal
dialysis patients. Am J Kidney Dis 2001; 37:33742.
30. Hemmelgarn BR, Manns BJ, Quan H, Ghali WA. Adapting
the Charlson Comorbidity Index for use in patients with
ESRD. Am J Kidney Dis 2003; 42:12532.
31. van Manen J, Korevaar J, Dekker F, Boeschoten E, Bossuyt
P, Krediet R. How to adjust for comorbidity in survival studies in ESRD patients: a comparison of different indices.
Am J Kidney Dis 2002; 40:829.
32. Beddhu S, Bruns FJ, Saul M, Seddon P, Zeidel ML. A simple
comorbidity scale predicts clinical outcomes and costs in
dialysis patients. Am J Med 2000; 108:60913.
33. Chin M, Goldman L. Correlates of early hospital readmission or death in patients with congestive heart failure.
Am J Cardiol 1997; 79:16404.
34. Extermann M. Measuring comorbidity in older cancer patients. Eur J Cancer 2000; 36:45371.
35. Gabriel SE, Crowson CS, OFallon WM. A comparison of two
comorbidity instruments in arthritis. J Clin Epidemiol
1999; 52:113742.
36. Davies S, Russell L, Bryan J, Phillips L, Russell G. Comorbidity, urea kinetics and appetite in continuous ambulatory peritoneal dialysis patients: their interrelationship
and prediction of survival. Am J Kidney Dis 1995; 26:
35361.
37. Davies S, Phillips L, Naish P, Russell G. Quantifying comorbidity in peritoneal dialysis patients and its relationship
to other predictors of survival. Nephrol Dial Transplant
2002; 17:108592.
38. Jager K, Merkus M, Dekker F, Els W, Tijssen J, Stevens P,
et al. Mortality and technique failure in patients starting
chronic peritoneal dialysis treatment: results of the Netherlands Cooperative Study on The Adequacy of Dialysis.
Kidney Int 1999; 55:147685.
39. Wright L. Survival in patients with end-stage renal disease. Am J Kidney Dis 1991; 17:258.
40. Greenfield S, Blanco DM, Elashoff RM, Aronow HU. Development and testing of a new index of comorbidity [Abstract]. Clin Res 1987; 35:346A.
41. Miskulin D, Athienites N, Yan G, Martin A, Ornt D, Kusek J,
et al. Comorbidity assessment using the Index of Coexistent Diseases (ICED) in a multicenter clinical trial: the Hemodialysis (HEMO) Study. Kidney Int 2001; 60:1498510.

PDI

JULY 2005 VOL. 25, NO. 4

42. Levey AS, Martin AA, Miskulin DC, Athienites NV, Meyer
KB. Comorbidity assessment in ESRD: methods and results
in a national dialysis chain [Abstract]. J Am Soc Nephrol
1999; 10:171(A).
43. Miskulin D, Meyer K, Martin A, Fink N, Coresh J, Powe N,
et al. Baseline comorbidity and its change predict survival
in an incident dialysis population. Am J Kidney Dis 2003;
41:14961.
44. McClellan W, Anson D, Birkeli K, Tuttle E. Functional status and quality of life: predictors of early mortality among
patients entering treatment for end-stage renal disease.
J Clin Epidemiol 1991; 44:839.
45. Miskulin DC, Meyer KB, Athienites NV, Martin AA, Marsh
JV, Fink NF, et al. The contribution of medical conditions
and physical impairments to a comorbidity index: the
Index of Coexistent Disease (ICED) [Abstract]. J Am Soc
Nephrol 1999; 10:175(A).
46. Chanda S, Schulz J, Lawrence C, Greenwood R, Farrington
K. Is there a rationale for rationing chronic dialysis? A
hospital based cohort study of factors affecting survival
and morbidity. Br Med J 1999; 318:21723.
47. Hanley JA, McNeil BJ. The meaning and use of the area
under a receiver operating characteristic (ROC) curve.
Radiology 1983; 143:2936.
48. Miskulin D, Martin A, Brown R, Fink N, Coresh J, Powe N,
et al. Predicting one-year mortality in an outpatient hemodialysis population: a comparison of comorbidity instruments. Nephrol Dial Transplant 2004; 19:41320.
49. Cook D. Performance of the APACHE III models in an Australian ICU. Chest 2000; 118:17328.
50. Knaus W, Wagner D, Draper E, Zimmerman J, Bergner M,
Bastos P, et al. The APACHE III prognostic system: risk prediction of hospital mortality for critically ill hospitalized
adults. Chest 1991; 100:161936.
51. Spertus JA, Winder JA, Dewhurst TA, Deyo RA, Prodzinski
J, McDonell M, et al. Development and evaluation of the
Seattle Angina Questionnaire: a new functional status
measure for coronary artery disease. J Am Coll Cardiol
1995; 25:33341.
52. Greenfield S, Sullivan L, Dukes KA, Silliman R, DAgostino
R, Kaplan SH. Development and testing of a new measure
of case mix for use in office practice. Med Care 1995; 33
(4 Suppl):AS47AS55.
53. DeOreo PB. Hemodialysis patient-assessed functional
health status predicts continued survival, hospitalization,
and dialysis-associated compliance. Am J Kidney Dis 1997;
30:20412.
54. Mapes D, Lopes A, Satayathum S, McCullough K, Goodkin
D, Locatelli F, et al. Health-related quality of life as a predictor of mortality and hospitalization: the Dialysis Outcomes and Practice Patterns Study (DOPPS). Kidney Int
2003; 64:33949.
55. Zager P, Nikolic J, Brown R, Campbell M, Hunt W, Peterson
D, et al. U curve association of blood pressure and mortality in hemodialysis patients. Kidney Int 1998; 54:5619.
56. Fleischmann E, Teal N, Dudley J, May W, Bower J,

MEASURING COMORBIDITY IN DIALYSIS PATIENTS

57.

58.

59.

60.

61.

62.

63.
64.

65.

66.

67.

68.

69.

70.

71.

72.

Salahudeen A. Influence of excess weight on mortality and

hospital stay in 1346 hemodialysis patients. Kidney Int
1999; 55:15607.
Leavey S, Strawderman R, Jones C, Port F, Held P. Simple
nutritional indicators as independent predictors of mortality in hemodialysis patients. Am J Kidney Dis 1998; 31:
9971006.
Block G, Hulbert-Shearon T, Levin N, Port F. Association
of serum phosphorus and calcium x phosphate product
with mortality risk in chronic hemodialysis patients: a
national study. Am J Kidney Dis 1998; 31:60417.
Kalantar-Zadeh K, Kopple J, Block G, Humphreys M. A
malnutrition-inflammation score is correlated with morbidity and mortality in maintenance hemodialysis patients. Am J Kidney Dis 2001; 38:125163.
Zimmermann J, Herrlinger S, Pruy A, Metzger T, Wanner
C. Inflammation enhances cardiovascular risk and mortality in hemodialysis. Kidney Int 1999; 55:64858.
Myers G, Rifai N, Russell P, Roberts W, Alexander R, Biasucci
L, et al. CDC/AHA workshop on markers of inflammation
and cardiovascular disease. Circulation 2004; 110:e5459.
Miskulin D, Coresh J, Klag M, Levin N, Fink N, Powe N, et al.
Can serum CRP and IL-6 levels replace comorbidity in predicting mortality in dialysis patients [Abstract]? J Am Soc
Nephrol 2004; 15:382A.
Elwood M. Critical Appraisal of Epidemiological Studies
and Clinical Trials. 2nd ed. New York, NY: Oxford; 1998.
Greene T, Beck GJ, Gassman JJ, Gotch FA, Kusek JW, Levey
AS, et al. Design and statistical issues of the Hemodialysis (HEMO) Study. Control Clin Trials 2000; 21:50225.
Liu Y, Coresh J, Eustace J, Longenecker J, Jaar B, Fink N,
et al. Association between cholesterol level and mortality in dialysis patients. JAMA 2004; 291:4519.
Lowrie EG, Lew NL. Death risk in hemodialysis patients:
the predictive value of commonly measured variables and
an evaluation of death rate differences between facilities.
Am J Kidney Dis 1990; 15:45882.
Iseki K, Yamazato M, Tozawa M, Takishita S. Hypocholesterolemia is a significant predictor of death in a cohort of
chronic hemodialysis patients. Kidney Int 2002; 62:
188793.
Coresh J, Longenecker J, Miller E. Epidemiology of cardiovascular risk factors in chronic renal disease. J Am Soc
Nephrol 1998; 9(Suppl 1):S2430.
Kalantar-Zadeh K, Block F, Humphreys M, Kopple J. Reverse epidemiology of cardiovascular risk factors in maintenance dialysis patients. Kidney Int 2003; 63:793808.
Fleischmann E, Bower J, Salahudeen A. Risk factor paradox in hemodialysis: better nutrition as a partial explanation. ASAIO J 2001; 47:7481.
Leavey SF, McCullough K, Hecking E, Goodkin D, Port F,
Young E. Body mass index and mortality in healthier as
compared with sicker haemodialysis patients: results
from the Dialysis Outcomes and Practice Patterns Study
(DOPPS). Nephrol Dial Transplant 2001; 16:238694.
Stack A, Molony D, Rahman N, Dosekun A, Murthy B.
331

MISKULIN

73.

74.

75.

76.

77.
78.

332

Impact of dialysis modality on survival of new ESRD patients with congestive heart failure in the United States.
Kidney Int 2003; 64:10719.
Stack A, Murthy B, Molony D. Survival differences between
peritoneal dialysis and hemodialysis among large ESRD
patients in the United States. Kidney Int 2004; 65:
2398408.
Centers for Medicare & Medicaid Services. 2001 ESRD Clinical Performance Measures Project. Am J Kidney Dis 2002;
39(5 Suppl 3):S198.
Greenfield S, Nelson EC, Zubkoff M, Manning W, Rogers W,
Kravitz RL, et al. Variations in resource utilization among
medical specialties and systems of care. JAMA 1992; 267:
162430.
Lacson Jr E, Teng M, Lazarus J, Lew N, Lowrie E, Owen W.
Limitations of the facility-specific standardization mortality ratio for profiling health care quality in dialysis
units. Am J Kidney Dis 2001; 37:26775.
Chertow GM. Leveling the paying field in end-stage renal
disease. Am J Med 2000; 108:6668.
Tonelli M, Gill J, Pandeya S, Bohm C, Levin A, Kiberd B.
Barriers to blood pressure control and angiotensin enzyme

JULY 2005 VOL. 25, NO. 4

79.

80.

81.

82.

83.

PDI

inhibitor use in Canadian patients with chronic renal insufficiency. Nephrol Dial Transplant 2002; 17:142633.
Manley H, Garvin C, Drayer D, Reid G, Bender W, Neufeld T,
et al. Medication prescribing patterns in ambulatory hemodialysis patients: comparisons of USRDS to a large notfor-profit dialysis provider. Nephrol Dial Transplant 2004;
19:18428.
National Kidney Foundation. K/DOQI clinical practice
guidelines for managing dyslipidemias in chronic kidney
disease. Am J Transplant 2004; 4(Suppl 7):753.
Nissenson A, Collins J, Dickmeyer J, Litchfield T, Mattern
W, McMahill C, et al. Evaluation of disease-state management of dialysis patients. Am J Kidney Dis 2001; 37:
93844.
Manley H, Drayer D, Muther R, Hebbar S, Miskulin D. Impact of multidisciplinary team on cardiovascular risk (CV)
reduction in hemodialysis (HD) patients: Dialysis Risk Factor Intervention Trial [Abstract]. J Am Soc Nephrol 2004;
15:191A.
Manley H, Drayer D. Clinical pharmacy interventions reduce ambulatory hemodialysis (HD) patients cardiovascular risk [Abstract]. J Am Soc Nephrol 2004; 15:3A.