Rubin HR, Fink NE, Plantinga LC, Sadler JH, Kliger AS, Powe NR. Patient Ratings of Dialysis Care With Peritoneal Dialysis vs Hemodialysis. JAMA. 2004;291(6):697–703. doi:10.1001/jama.291.6.697
Author Affiliations: Department of Medicine, The Johns Hopkins School of Medicine, Baltimore, Md (Drs Rubin and Powe and Mss Fink and Plantinga); Department of Health Policy and Management (Drs Rubin and Powe) and Department of Epidemiology (Drs Rubin and Powe and Ms Fink), The Johns Hopkins Bloomberg School of Public Health, Baltimore, Md; Independent Dialysis Foundation, Baltimore, Md (Dr Sadler); and Department of Medicine, Hospital of St Raphael and Department of Medicine, Yale University, New Haven, Conn (Dr Kliger).
Context In light of conflicting evidence of differential effects of dialysis
modality on survival, patient experience becomes a more important consideration
in choosing between hemodialysis and peritoneal dialysis.
Objective To compare patient satisfaction with hemodialysis and peritoneal dialysis
in a cohort of patients who have recently begun dialysis.
Design and Setting Cross-sectional survey at enrollment in a prospective inception cohort
study of patients who recently started dialysis at 37 dialysis centers participating
in the Choices for Healthy Outcomes in Caring for End-stage Renal Disease
(CHOICE) study, a national multicenter study of dialysis outcomes, from October
1995 to June 1998.
Patients Of 736 enrolled incident dialysis patients, 656 (89%) returned a satisfaction
questionnaire after an average of 7 weeks of dialysis.
Main Outcome Measure Data collected from a patient-administered questionnaire including 3
overall ratings and 20 items rating specific aspects of dialysis care.
Results Patients receiving peritoneal dialysis were much more likely than those
receiving hemodialysis to give excellent ratings of dialysis care overall
(85% vs 56%, respectively; relative probability, 1.46 [95% confidence interval,
1.31-1.57]) and significantly more likely to give excellent ratings for each
specific aspect of care rated. The 3 items with the greatest differences were
in the domain of information provided (average of information items: peritoneal
dialysis [69% excellent] vs hemodialysis [30% excellent]). The smallest differences
were in ratings of accuracy of information from the nephrologist, response
to pain, amount of fluid removed, and staff availability in an emergency.
Adjustment for patient age, race, education, health status, marital status,
employment status, distance from the center, and time since starting dialysis
did not reduce the differences between peritoneal dialysis and hemodialysis
Conclusions After several weeks of initiating dialysis, patients receiving peritoneal
dialysis rated their care higher than those receiving hemodialysis. These
findings indicate that clinicians should give patients more information about
the option of peritoneal dialysis.
Nearly 100 000 patients with incident end-stage renal disease (ESRD)
in the United States must choose between hemodialysis and peritoneal dialysis
each year in light of a limited number of donor kidneys for transplantation.1 Since the 1970s, patients have had the choice of these
2 modalities. Hemodialysis is usually performed in an outpatient dialysis
facility 3 times a week for 3 to 4 hours, where trained nurses and technicians
carry out the prescribed treatment using a dialysis machine under the direction
of a physician.
In contrast, peritoneal dialysis is most commonly performed every day
at home by the patient after receiving training by dialysis facility staff.
Physicians can prescribe peritoneal dialysis using manual exchanges of dialysate
fluid (continuous ambulatory peritoneal dialysis [CAPD]) 4 to 5 times a day
or automated exchanges in which a cycler is used to fill and drain the peritoneum
with dialysate fluid usually at night while asleep (continuous cycling peritoneal
dialysis). Because peritoneal dialysis is done at home, patients do not have
to visit the facility as often (eg, once per month), although they may be
in contact with dialysis facility staff. Hemodialysis requires establishing
access to the vasculature through an arteriovenous fistula, graft, or catheter,
whereas peritoneal dialysis requires establishing access to the peritoneal
cavity using a peritoneal catheter.
Research has yielded conflicting results about mortality differences
between the 2 modalities; studies with better adjustment for comorbidities
have found either no mortality differences or better survival with peritoneal
dialysis, especially in the first 2 years after starting dialysis.2- 4 On a per-treatment basis,
both hemodialysis and peritoneal dialysis are efficacious in removing solutes
and water depending on the clearance characteristics of the dialysis membrane
(artificial or peritoneum, respectively) and the time allowed for exchange
across the membrane. In the long term, peritoneal dialysis efficacy can be
limited by recurrent peritonitis and the loss of peritoneal clearance or residual
kidney function, whereas hemodialysis efficacy may be limited by difficulty
in obtaining or maintaining vascular access and poor hemodynamic tolerance.
When treatments provide equivalent survival benefits and clinical efficacy,
patients' views of their treatment become more important in choosing among
options. Patients' views are influenced by factors besides effects on survival.
Important factors contributing to patient choice include whether they or the
staff do the treatment, how often they receive dialysis, and whether there
is flexibility allowing them control over their schedules.5 Factors
that influence patient evaluations of dialysis care include explanations received,
availability of nephrologists, amount of fluid removed, and availability of
Many physicians believe that peritoneal dialysis is underused in the
United States,8 which may indicate that nephrologists
are weighing factors besides mortality rates. Because survival differences
are uncertain, both patients and clinicians assisting them in selecting the
type of dialysis could benefit from knowing how patients receiving different
modalities evaluate them. However, few studies have described patients' experiences
and evaluations of the 2 modalities and have not directly compared ratings
of the 2 modalities by similar patients.5 Therefore,
we conducted a study to compare evaluations of dialysis care by peritoneal
and hemodialysis patients at the same dialysis centers after several weeks
of initiating dialysis, taking into account patient case mix.
The Choices for Healthy Outcomes in Caring for End-stage Renal Disease
(CHOICE) study is a prospective cohort study to investigate treatment choices
and outcomes of dialysis care among patients beginning dialysis .9,10 Eligibility criteria for enrollment
included initiation of chronic outpatient dialysis in the preceding 3 months,
ability to give informed consent for participation, age older than 17 years,
and the ability to speak English or Spanish. The Johns Hopkins University
School of Medicine institutional review board and the review boards for the
clinical centers approved the study protocol. All patients provided written
From October 1995 to June 1998, 1041 participants were enrolled at 81
dialysis centers associated with Dialysis Clinic, Inc (Nashville, Tenn; n
= 923), New Haven CAPD (New Haven, Conn; n = 86), or Saint Raphael's Hospital
(New Haven, Conn; n = 32). As part of the baseline evaluation, we obtained
patient ratings of their dialysis care. Of the 81 participating dialysis centers
in the CHOICE study, 37 centers in 14 states offered both hemodialysis and
peritoneal dialysis and were included in this comparison.
Dialysis modality at baseline was defined as the modality at 4 weeks
after enrollment in the study. All forms of peritoneal dialysis were combined
as a single category. Measures included a validated, chart-based comorbidity
and severity of illness measure, the Index of Coexistent Disease (ICED),11 as well as self-administered questionnaires containing
demographic and health status measures and evaluations of care.6,9- 11 Health
status measures included the Medical Outcomes Study 36-Item Short-Form Health
Survey (MOS SF-36),12,13 incorporating
both physical and mental health status domains, as well as scales developed
specifically to supplement the MOS SF-36 for ESRD patients that have been
described elsewhere.10 The developmental and
final versions of the questionnaire measuring patient evaluations and satisfaction
of care have been described elsewhere.6 Briefly,
the 23-item instrument was developed using focus groups of patients receiving
hemodialysis and peritoneal dialysis. From the focus groups, more than 100
items were prioritized by a sample of 86 dialysis patients to construct the
instrument. Domains included nephrologists' availability, technical quality,
interpersonal treatment, information, and coordination; nurses' and dialysis
center staff's availability, technical quality, interpersonal treatment, response
to pain, and information; and social worker availability. Three global questions
on overall quality were also included. For each question the response set
consisted of 5 categories: poor, fair, good, very good, and excellent.
The printed questionnaire in English or Spanish including health status
and patient evaluation measures was given to patients to complete during their
enrollment visit. Patient names were not written on the survey and an envelope
was provided for the patients to seal their completed surveys to protect the
confidentiality of their response.
Differential completion by dialysis modality could bias results if questionnaire
respondents were more or less satisfied than nonrespondents. Therefore, we
compared demographic and clinical characteristics of respondents and nonrespondents.
In bivariate analyses, we compared demographic and clinical characteristics
of peritoneal dialysis and hemodialysis respondents using differences in proportions
with 95% confidence intervals (CIs) for categorical variables and differences
in means with 95% CIs for continuous variables.
In bivariate analyses, the proportions of excellent ratings of each
item were compared for patients receiving hemodialysis and peritoneal dialysis.
We also examined the relationship of other patient characteristics to the
proportion of excellent ratings to explore potential confounders of the relationship
between modality and ratings of care. In multivariable analyses, we used multiple
logistic regression analysis to examine the presence, strength, and independence
of the relationship between modality and patient ratings of their care. We
used the generalized estimating equation model to take into account potential
dialysis center effects on patient ratings of care.14 Patient
age, sex, race, educational level, marital status, occupational status, comorbidity
(ICED), self-rated health status (the 2 component subscales of the MOS SF-36,
the Mental Component Score [MCS] and the Physical Component Score [PCS]),
distance from dialysis center, and time since starting dialysis were used
as adjustment variables. Adjusted probabilities of excellent ratings for peritoneal
dialysis compared with hemodialysis were calculated from the logistic regression
analyses, adjusting to distribution of patient characteristics present in
the hemodialysis sample.
We performed sensitivity analyses under various conditions to test the
robustness of our analysis: (1) using propensity scores calculated from all
independent variables rather than all variables to calculate adjusted probabilities
of an excellent rating; (2) grouping very good and excellent vs all other
scores, rather than excellent vs all other scores; (3) using ordinal logistic
regression analyses rather than logistic regression; and (4) including all
clinics regardless of modalities offered instead of only those that offered
both peritoneal dialysis and hemodialysis.
We also conducted these analyses within various subgroups to determine
if the relationships held for all types of patients. Subgroups were created
dichotomizing each patient characteristic in the model and included the following:
those younger than 65 years, those 65 years or older, males, females, whites,
nonwhites, those employed, those not employed, those with MOS SF-36 MCS scores
less than 47.2 (median value), those with MCS scores of 47.2 or higher, those
with lower ICED scores (0-1), those with higher ICED scores (2-3), those who
lived at least 48 km from the dialysis center, those who lived less than 30
miles from the dialysis center, those who had been receiving dialysis less
than 43 days, and those who had been receiving dialysis for 43 days or more.
The level of significance was P = .05 and Stata software
version 7.0 (Stata Corp, College Station, Tex) was used for all analyses.
Among the 736 patients receiving both dialysis modalities, 95% agreed
to complete surveys and 656 (89%) responded, 521 (71%) with complete questionnaires
and an additional 135 (18%) with partially complete questionnaires. These
responding patients had been receiving dialysis an average of 7 weeks (range,
0-26 weeks) at the time they completed the baseline satisfaction questionnaire.
Among the 480 eligible hemodialysis patients, 336 (92%) responded; 107 (22%
of those eligible) returned partially complete questionnaires. There were
no statistically significant differences between hemodialysis respondents
and nonrespondents, although there was a marginally significant finding that
respondents were more likely to be high school graduates (68.0% of respondents
vs 51.9% of nonrespondents, P = .08).
Among 256 peritoneal dialysis patients, 185 (85%) responded, 28 (11%)
with partially complete questionnaires. Among peritoneal dialysis patients,
respondents were more likely than nonrespondents to be white (80.2% vs 60.5%, P = .004), and there were marginally significant differences
in mean age (respondents, 54.6 years; nonrespondents, 49.9 years; P = .06) and ICED score (score of 2-3: respondents, 51.6%; nonrespondents,
37.2%; P = .08).
A greater proportion of peritoneal dialysis patients than hemodialysis
patients were white, married, high school graduates, working full- or part-time,
and living more than 30 miles from the dialysis center (Table 1). Although self-reported health status scores were similar
among patients receiving the 2 modalities, patients receiving peritoneal dialysis
were much less likely than hemodialysis patients to have comorbidity, as measured
by the ICED score, at the start of dialysis.
Bivariate analyses of relationships of patient characteristics to ratings
showed that age, education, marital status, employment, time since starting
dialysis, and self-reported health status scores were not statistically significantly
associated with the likelihood of giving excellent overall ratings. Whites
were more likely to rate care excellent overall than nonwhites were (69% vs
57%, respectively, P = .003). Those who lived 48
km or more from dialysis centers were also more likely to rate care excellent
overall than those who lived less than 48 km away (75% vs 63%, respectively, P = .03). Those with less vs more comorbidity/disease severity
(73% vs 61%, respectively, P = .002) were more likely
to rate care excellent overall.
There were fewer associations between patients' characteristics and
their indicating that they would definitely recommend their dialysis centers
to others. A greater percentage of working than nonworking respondents (87%
vs 78% respectively, P = .04) would definitely recommend
their center. Those who had started dialysis more recently (<43 days) were
more likely to report that they would recommend their center to others than
those who had been receiving dialysis for a longer time (70% vs 61%, respectively, P = .03). In addition, 86% of those with better than average
mental health (MOS SF-36 MCS scores ≥47.2) vs 74% of those with poorer
mental health (P<.001) indicated that they would
definitely recommend their center to others.
The best rated (unadjusted) items overall were similar for both modalities:
"caring and concern of nurses," "caring and concern of dialysis center staff,"
"attention to cleanliness," "response to pain," and "availability in emergencies"
(Table 2). Four of the 5 worst-rated
items overall were the same for peritoneal dialysis and hemodialysis patients:
"coordination among nephrologists and other doctors," "how often the nephrologist
sees you," "the amount of fluid removed during dialysis," and "the accuracy
of nephrologists' information." For peritoneal dialysis patients, the other
item rated in the 5 worst (16/20) was "ease of reaching the nephrologist,"
whereas this item ranked 14/20 in hemodialysis patients' ratings. For hemodialysis
patients, the other item in the 5 worst (17/20) was "the amount of information
given to help choose hemodialysis or peritoneal dialysis"; this item was ranked
8/20 by peritoneal dialysis patients.
Peritoneal dialysis patients gave much higher ratings of care than hemodialysis
patients (Table 2). Overall care
was rated excellent by 85% of peritoneal dialysis vs 56% of hemodialysis patients.
The specific items with the largest differences between the 2 modalities were
"information given to help choose modality" and "the amount of dialysis information
After adjustment to reflect the patient characteristics of the hemodialysis
population including age, sex, education, employment, health status, comorbidity,
time since beginning dialysis, distance from dialysis center, and dialysis
center, a higher adjusted percentage of patients receiving peritoneal dialysis
gave an excellent rating than those receiving hemodialysis for every one of
the 23 items on the survey (Table 2).
The items with the greatest adjusted ratio of probabilities of an excellent
rating for peritoneal dialysis relative to hemodialysis patients concerned
information given by dialysis staff. For the 3 items pertaining to information,
on average, peritoneal dialysis patients would be 2.2 times as likely to give
an excellent rating as similar hemodialysis patients. Items rated most similarly
by peritoneal dialysis and hemodialysis patients were "response to pain,"
"staff availability in emergency," and "caring of dialysis staff."
No patient demographic or health status variables were statistically
significantly associated with overall excellent ratings of the quality of
dialysis care, after adjustment for dialysis modality and center identity
(data not shown). For the 22 other rating items, there also were few such
associations; higher mental health scores (MOS SF-36 MCS) were associated
with better ratings on 12 items, younger age on 8 items, white race on 3 items,
high school graduate on 2 items, less time since starting dialysis on 2 items,
and male sex on 1 item.
Results of the 4 sensitivity analyses conducted were all consistent
with the original analysis. For example, using all clinics rather than only
clinics that provided both modalities, the relative probability of an excellent
overall rating for those receiving peritoneal dialysis relative to hemodialysis
was 1.40 (95% CI, 1.28-1.50), compared with the original analysis result of
1.46. Using propensity scoring and ordinal logistic regression, the odds ratios
for excellent overall ratings for peritoneal relative to hemodialysis patients
were also almost identical to those derived in the original analyses (3.59
and 3.55, respectively, vs 3.61). Collapsing very good and excellent ratings
together provided a less sensitive measure of differences in opportunities
for improvement, but this grouping maintained the higher relative probability
of positive ratings among peritoneal patients than among hemodialysis patients
(1.13 [95% CI, 1.05-1.17]).
Adjusted percentages of excellent overall quality ratings were higher
for peritoneal dialysis than hemodialysis patients in all 16 subgroups examined,
ranging from 1.14 to 1.79. This relative probability of an excellent rating
was statistically significantly greater than 1 in all but 2 of the 16 subgroups:
employed patients and those who lived 48 km or more from the dialysis center.
These 2 subgroups each had fewer than 90 patients and limited power.
Patients receiving peritoneal dialysis were 1.5 times as likely to rate
their dialysis care excellent overall than were patients receiving hemodialysis.
This was true for patients with different demographic and health status characteristics.
In addition, peritoneal dialysis patients more often rated their care for
many specific aspects more highly than hemodialysis patients did. This finding
is important in light of the diminishing proportion of dialysis patients in
the United States who are beginning renal replacement therapy with peritoneal
This study highlights the need to improve information provided to patients,
particularly information to help patients choose between hemodialysis and
peritoneal dialysis. Information about choosing a dialysis modality was the
single item rated most differently by patients in the 2 groups and was rated
much higher by peritoneal dialysis patients. This suggests that more informed
patients choose peritoneal dialysis more often, as has also been noted in
the United Kingdom.15
We had hypothesized that peritoneal dialysis patients would rate some
items less favorably than hemodialysis patients would, for example, ease of
reaching their nephrologist. Peritoneal dialysis patients who visit the dialysis
center less frequently must often make contact with their physician by telephone,
whereas hemodialysis patients who visit a facility typically 3 times a week
might be expected to have more opportunity to reach their physician in person.
However, this item ranked in the bottom 5 ratings for peritoneal dialysis
patients but was still rated higher by peritoneal dialysis than hemodialysis
patients. Thus, despite seeing the same nephrologists and social workers in
most cases and having fewer in-person contacts with them, peritoneal dialysis
patients rated the same clinicians higher than their counterparts receiving
hemodialysis at the same center. This may be because patients value availability
whether by telephone or in person, but may also reflect intensive early interactions
with nephrologists and social workers that diminish over time. At least in
this early period, it is likely that clinicians do give better care to peritoneal
dialysis patients in the dimensions of information and interpersonal treatment
that patients value most highly, perhaps because patients must be better educated
and supported by their clinicians to learn and accomplish home dialysis.
Dialysis patients often look to their nephrologists as their principal
care provider, yet their substantial and growing burden of comorbidity necessarily
involves the input of other physicians such as surgeons and radiologists for
vascular access, cardiologists for the common occurrence of cardiovascular
disease, emergency physicians for urgent after hours care, and sometimes primary
care physicians for routine screening and preventive services. Our results
suggest that coordination between nephrologists and other physicians needs
improvement for all patients. This is a concern particularly for hemodialysis
patients for whom it was the lowest-rated item on the survey. Coordination
of care involves planning care; assigning and communicating clear responsibility
for roles, tasks, and priorities by all involved persons; and sharing information
across time, persons, and settings. Dialysis patients have a chronic illness
for which they require evaluation and treatment by many physicians across
multiple settings including hospitals, physicians' offices, and dialysis facilities.
Coordination is a challenge in light of the complexity of this illness and
payment policy that may foster compartmentalization of care. To date, although
the importance of coordinated care has been acknowledged,16,17 system
changes to address this concern have been unsuccessful.18- 21
In contrast to the pattern often seen in patient evaluations of inpatient
care,22 nurses and center staff were rated
much higher than physicians by both groups. This is probably a result of the
regular involvement of staff in the care of dialysis patients at freestanding,
often community-based, facilities and indicates how dependent patients may
be on such staff. Current nursing shortages could pose a threat to the satisfaction
of a growing number of dialysis patients in the United States.
Comparisons of dialysis modalities are difficult because randomized
controlled trials in which clinicians and patients can be masked to treatment
group are not feasible. Therefore, careful attention to case-mix adjustment
is important. We found many differences between patients. In particular, the
measure of the burden of comorbid diseases and their severity, the ICED, was
greater in hemodialysis patients than peritoneal dialysis patients. We used
multivariable methods to assess the effect of treatment modality on patient
ratings independent of sociodemographics or health status. In addition, stratified
analyses showed similar findings to our overall results in almost all of the
Limitations of this study deserve comment. First, surveys can be biased
if there is differential nonresponse in the groups being compared. The adjusted
and stratified analyses make it less likely that response bias influenced
our results, but this concern cannot be entirely dismissed. Peritoneal dialysis
patients choose an active form of therapy and may be those with a greater
sense of control over their lives preceding their choice of dialysis. This
may translate to greater satisfaction with all their choices. However, one
of the largest differences observed between patients receiving the 2 modalities
was that peritoneal dialysis patients gave much higher ratings of the information
they were given about the choice of modalities. This suggests that there is
an opportunity to inform and encourage more patients to begin with peritoneal
dialysis rather than hemodialysis, which may translate into patients feeling
more in control of their lives.
Second, this study determined patients' evaluations of dialysis care
after an average of several weeks of dialysis. Ratings during this initial
period are important because they could guide initial decisions about modality.
Having highly rated care early in treatment can ease the lifestyle adjustments
that patients often must make in living with a severe chronic illness and
in their transition to dialysis care. However, it will be important to compare
ratings of the 2 modalities later in dialysis once patients have had a chance
to experience some of the complications of their chosen modality, and the
data presented here do not address these later impressions.
Third, patient ratings are but one aspect of the quality of dialysis
care; quality of life and measures of clinical performance established by
health professionals are important supplements to this study. The importance
of our results is predicated on previous research indicating that the modalities
are comparable in clinical effectiveness as described earlier. Patient ratings,
however, do predict change in clinicians23,24 and
evaluate dimensions of care such as communication and interpersonal treatment
that are not evaluated by any other method.
Fourth, although we adjusted our results to the hemodialysis population,
inferences from observational studies must be made cautiously because of the
possibility of selection and nonresponse biases referred to earlier.
Finally, our study was not a pure comparison of hemodialysis and peritoneal
dialysis modalities, because what we have really compared is center-based
hemodialysis with home-based peritoneal dialysis. The advantages of home-based
peritoneal dialysis over center-based hemodialysis may also apply to home
or limited care hemodialysis, and this is worthy of future study.
The results of this study may be of particular interest because peritoneal
dialysis may be less costly than hemodialysis when all societal costs are
considered. This might imply that the Medicare program in the United States
could curtail costs substantially if more patients were started on peritoneal
dialysis, while aligning care better with consumer preferences. Medicare's
current ESRD payment system, which pays similarly for each modality,25 may not achieve this result. For facilities, the
components contributing to cost include capital (eg, facility and equipment),
labor (eg, nurses, technicians), and supplies (eg, dialysate fluids, tubing).
The capital and labor costs of hemodialysis are more expensive because patients
receive care in a treatment facility 3 times a week under the supervision
of staff. Medicare financial incentives currently may encourage clinicians
to choose hemodialysis because once substantial investment in a hemodialysis
facility has been made, the marginal costs of treating an additional patient
are likely lower for a new hemodialysis patient than for a new peritoneal
dialysis patient. Recent proposed changes in physician payment policy for
home and in-center dialysis would pay physicians by the number of in-person
contacts or visits and thus may create the unintended consequence of an even
greater disincentive for physicians to provide home dialysis treatment, particularly
in rural areas.26
Thus, changes in current and planned payment policies would need to
occur for the Medicare program to realize the potential cost savings of more
patients choosing peritoneal dialysis. On the other hand, in revising payment
policies, care must be taken not to penalize physicians for fully informed
patients who choose hemodialysis or for choosing hemodialysis when peritoneal
dialysis is contraindicated.
In conclusion, our results that peritoneal dialysis patients rate their
care more highly than hemodialysis patients suggest that nephrologists and
primary care physicians should give greater consideration to peritoneal dialysis
when patients are eligible for either modality, especially in light of no
clear superiority in survival. More thorough information about choice of modality
prior to the start of renal replacement therapy may lead more patients to
choose peritoneal dialysis and lead to better patient satisfaction.