Sehgal AR, Leon JB, Siminoff LA, Singer ME, Bunosky LM, Cebul RD. Improving the Quality of Hemodialysis TreatmentA Community-Based Randomized Controlled Trial to Overcome Patient-Specific Barriers. JAMA. 2002;287(15):1961-1967. doi:10.1001/jama.287.15.1961
Author Affiliations: Division of Nephrology, MetroHealth Medical Center (Dr Sehgal, Mss Leon and Bunosky); Center for Health Care Research and Policy, MetroHealth Medical Center (Drs Sehgal, Singer, and Cebul); and Department of Medicine, (Drs Sehgal, Siminoff, and Cebul), Department of Epidemiology and Biostatistics (Drs Sehgal, Singer, and Cebul); and Center for Biomedical Ethics, Case Western Reserve University (Drs Sehgal and Siminoff, Cleveland, Ohio).
Context Mortality rates among US hemodialysis patients are the highest in the
industrialized world at 23% per year. Measures of dialysis dose (Kt/V) correspond
strongly with survival and are inadequate in one sixth of patients. Inadequate
dialysis is also associated with increased hospitalizations and high inpatient
costs. Our previous work identified 3 barriers to adequate hemodialysis: dialysis
underprescription, catheter use, and shortened treatment time.
Objective To determine the effect of a tailored intervention on adequacy of hemodialysis.
Design and Setting Community-based randomized controlled trial with recruitment from April
1999 to June 2000 at 29 hemodialysis facilities in northeast Ohio.
Participants Forty-four nephrologists and their 169 randomly selected adult patients
receiving inadequate hemodialysis.
Intervention Nephrologists were randomly assigned to an intervention (n = 21) or
control (n = 23) group. For patients in the intervention group (n = 85), depending
on the barrier(s) present, a study coordinator gave nephrologists recommendations
about optimizing dialysis prescriptions, expedited conversion of catheters
to surgically created grafts or fistulas, and educated patients about the
importance of compliance with treatment time. Patients in the control group
(n = 84) continued to receive usual care.
Main Outcome Measures Changes in Kt/V and specific barriers after 6 months.
Results At baseline, intervention and control patients had similar Kt/V measurements,
specific barriers, and demographic and medical characteristics. After 6 months,
intervention patients had 2-fold larger increases in Kt/V compared with control
patients (+0.20 vs +0.10; P<.001) and were more
likely to achieve their facility Kt/V goal (62% vs 42%; P = .01). Intervention patients also had nearly 3-fold larger increases
in dialysis prescription (+0.16 vs +0.06; P<.001)
and were 4 times more likely to change from use of catheters to use of fistulas/grafts
(28% vs 7%; P = .04).
Conclusions An intervention tailored to patient-specific barriers resulted in increased
hemodialysis dose. Extending this approach to the 33 000 persons in the
United States receiving inadequate hemodialysis may substantially enhance
patient survival, diminish hospitalizations, and decrease inpatient expenditures.
Of the 200 000 Americans receiving chronic hemodialysis treatment
for end-stage renal disease, approximately 33 000 receive an inadequate
dialysis dose.1,2 This contributes
to dialysis patient mortality rates that are the highest in the industrialized
world, at 23% per year.2 Inadequate dialysis
dose is also associated with increased hospitalizations and high inpatient
costs.3 As a result, hemodialysis patients
represent only 0.5% of Medicare beneficiaries but account for 5% of Medicare
Dose of dialysis is quantified by the parameter Kt/V, a measure of urea
(K) removal during treatment (t) that is derived from a logarithmic transformation
of the percentage reduction in urea concentration.5
For example, a 65% reduction in blood urea nitrogen (BUN) from 80 mg/dL (29
mmol/L) at the beginning of dialysis to 28 mg/dL (10 mmol/L) at the end of
dialysis corresponds approximately to a Kt/V of 1.20. Larger percentage reductions
in urea concentration result in a higher Kt/V.6
While optimal levels remain to be established, doses of dialysis represented
by Kt/V measurements less than 1.20 are widely considered inadequate.7 Below this level, each 0.10 decrease in Kt/V is estimated
to increase the relative risk of death by 7%.8
Recent work suggests that even higher Kt/Vs are beneficial, and many providers
now target a Kt/V of 1.30 or 1.40.9
We reasoned that efforts to improve the quality of hemodialysis treatment
should be informed by an understanding of patient-specific barriers. In fact,
efforts toward quality improvement in any area typically begin with literature
review and brainstorming sessions to identify potential barriers to optimal
outcomes.10 However, the number of potential
barriers identified by such methods is often very large, and it is difficult
to know which barriers should be the focus of initial intervention. In this
case, potential barriers to adequate hemodialysis include hypotension, comorbid
conditions, noncompliance, poor vascular access, underprescription, reuse
of dialyzers, and clotting.11
We also reasoned that promising barriers for intervention should have
4 features. First, a promising barrier should be independently associated
with a poor outcome. Second, a promising barrier should be frequently present
since uncommon barriers may be less fruitful areas for initial intervention.
Third, a promising barrier should have a large effect on the outcome of interest.
Fourth, a promising barrier should be modifiable.12
In a previous investigation of 721 randomly selected patients, we identified
3 potential barriers with all 4 features: underprescription of dialysis by
the nephrologist, use of catheters for vascular access (as opposed to surgically
created fistulas or grafts), and shortening of treatment time for the patient.
For example, 11% of patients had a catheter, catheter use was independently
associated with a 0.17 decrement in Kt/V, and catheters often can be replaced
by fistulas or grafts.11
Based on these findings, we hypothesized that overcoming the 3 barriers
would result in an increased dialysis dose. To test this hypothesis, we designed
a community-based randomized controlled trial targeting patients receiving
inadequate dialysis and their nephrologists.
We first invited 31 hemodialysis facilities in northeast Ohio to participate,
with 29 facilities agreeing. We then invited all 54 nephrologists practicing
at these facilities to participate, with 53 nephrologists agreeing. We used
a random-number generator to assign these nephrologists to an intervention
or control group (Figure 1), assigning
nephrologists rather than patients to prevent the possibility that a given
nephrologist may care for both intervention and control patients. This randomization
did not involve blocking or stratification.13
To limit the burden on individual nephrologists and facilities, a study coordinator
also used a random-number generator to select a maximum of 10 eligible patients
per physician and a maximum of 10% of the patients at each facility.
To determine eligibility, we identified patients receiving inadequate
dialysis by abstracting medical records to identify patients whose most recent
Kt/V and whose mean Kt/V for the previous 3 months were both less than the
monthly goal for their facility (calculated by applying a standard formula
to results of predialysis and postdialysis BUN determinations14).
This ensured that only patients with persistently low Kt/V values were identified.
Additional patient eligibility criteria were age 18 years or older and receiving
dialysis for at least 6 months. We excluded new patients because the first
several months of dialysis treatment often is a time of multiple changes in
dialysis prescription and vascular access. Subjects who declined to participate,
did not speak English, or were mentally impaired also were excluded. We obtained
informed consent from eligible patients, and they were each given $10 at the
beginning and again at the end of the trial to thank them for their participation.
This study was approved by the institutional review board of MetroHealth Medical
Center, Cleveland, Ohio.
We abstracted medical records of consenting patients to determine the
presence of 3 specific barriers to adequate dialysis: low prescription, catheter
use, and shortened treatment time.
While Kt/V (also called "delivered" Kt/V) is a measure of the amount
of dialysis actually received by a patient, it is also possible to calculate
the amount of dialysis prescribed for a patient (sometimes called "prescribed"
Kt/V).15 We calculated the prescribed Kt/V
based on manufacturers' specifications for the prescribed dialyzer at the
prescribed blood and dialysate flows, the prescribed treatment time, and patient
anthropometric volume.7,11,15- 17
Because the in vivo performance of dialyzers is somewhat less than manufacturers'
in vitro specifications, delivered Kt/V is, on average, 0.10 to 0.20 lower
than the prescribed Kt/V.6,7,11
Therefore, we subtracted 0.20 from the prescribed Kt/V and classified patients
as having a low prescription when this value was lower than the facility goal
for delivered Kt/V. Since a patient's body size is reflected in the calculation
of V, this method explicitly accounts for differences in the urea distribution
volume of large vs small patients.
We noted whether the most recent treatment involved use of a catheter
as opposed to a fistula or graft; we also noted prescribed vs actual treatment
time for all treatments that were accompanied by Kt/V measurements over the
previous 3 months. Patients who missed more than 5% of their prescribed treatment
time were classified as having a shortened treatment-time barrier.11
A study coordinator (J.B.L.) educated all intervention patients about
the meaning and importance of adequate dialysis dose. She then provided feedback
and recommendations to intervention patients (during a dialysis treatment)
and to nephrologists (in their offices). The information provided was based
on the specific barrier(s) present.
The study coordinator explained why the current prescription was too
low and how it could be optimized. If a low-efficiency dialyzer or submaximal
flows of blood or dialysate were being used, she recommended a higher-efficiency
dialyzer or higher flows of blood or dialysate. Otherwise, she recommended
an increased treatment duration. Prescribed Kt/V was calculated for the recommended
prescription changes to ensure that the new prescription was at least 0.20
above the facility goal for delivered Kt/V.
The study coordinator interviewed patients and nephrologists to ascertain
the reason for catheter use (eg, lack of vascular sites, frequent clotting,
patient unwillingness to undergo surgery to receive a fistula or graft, no
referral to vascular surgeon). If patients were unwilling to undergo surgery,
she educated patients about the negative impact of catheter use on dialysis
dose. If the patient had not been referred to a vascular surgeon, she encouraged
the nephrologist to do so. Otherwise, she recommended an increased dialysis
The study coordinator interviewed patients to ascertain reasons for
shortened treatment time (eg, dialysis-associated symptoms, transportation
problems, time conflicts related to work or family). She enlisted the aid
of the nephrologist to address patient symptoms and the facility's social
worker to address transportation problems and time conflicts.
During the 6-month duration of the trial, the study coordinator communicated
on a monthly basis with patients and nephrologists to reinforce the above
recommendations, monitor progress, and answer questions. If specific barriers
persisted, she noted whether this was due to a medical limitation, patient
refusal, physician refusal, or a facility-related impediment. The study coordinator
also assessed patient quality of life at the beginning and again at the end
of the trial using 7 subscales (general health, energy/fatigue, emotional
well-being, burden of kidney disease, dialysis-associated symptoms, sleep,
satisfaction with care) of the Kidney Disease Quality of Life instrument.18 Because the study coordinator carried out the intervention,
it was not possible for her to be blinded to subjects' assignment to intervention
vs control groups. There were no adverse events or side effects associated
with the intervention.
Control patients underwent quality-of-life assessments at the beginning
and again at the end of the trial. Otherwise, they continued to receive usual
care from their nephrologists. Neither control patients nor their nephrologists
received any feedback from study personnel.
All patients were recruited between April 1999 and June 2000 and followed
up for 6 months, or until they died or moved. During this interval, medical
records of intervention and control patients were abstracted on a monthly
basis to obtain data on Kt/V and specific barriers.
Primary outcomes were change in Kt/V and achievement of facility Kt/V
goal. To increase the precision of the estimates, the "final" Kt/V of each
patient was defined as the mean of all Kt/V measurements during months 4 through
6 of the trial while the "baseline" Kt/V was the mean of all Kt/V measurements
during the 3 months prior to subject enrollment.19
Change in Kt/V was calculated as the final Kt/V minus the baseline Kt/V. The
final Kt/V was used to determine if patients achieved their facility goal.
Because we expected it would take about 3 months for our intervention to have
an effect, patients who died or moved during months 1 through 3 of the trial
were not included in the main analyses. Secondary outcomes were (1) changes
in prescription, catheter use, and treatment time among patients who had these
barriers at baseline and (2) changes in quality of life among all patients.
Because nephrologists comprised the unit of randomization, our main
analyses account for the clustering of patients by nephrologist. Specifically,
we compared change in Kt/V for intervention patients vs control patients using
an adjusted t test that reflects the clustering of
patients by nephrologist.20 Similarly, we used
an adjusted χ2 test that reflects the clustering of patients
by nephrologist to compare the proportion of patients in each group that achieved
the facility Kt/V goal.21- 23
Changes in specific barriers and quality of life were examined using the χ2 test for dichotomous variables or the Mann-Whitney rank-sum test for
We calculated the sample size required to detect a 25% difference in
our dichotomous primary outcome (ie, 25% of control vs 50% of intervention
patients achieving the facility Kt/V goal). To detect this difference with
a power of 80% and an α of .05 requires 120 total subjects, or 2 to
3 patients per nephrologist.24 We then inflated
this estimate to account for possible nonindependence of subjects clustered
by nephrologist. We conservatively assumed that up to 75% of nephrologists
would have concordant results (ie, at the end of the trial, all of their patients
either will or will not have achieved the facility goal). This gives an inflation
factor of 1.33 and a final sample size requirement of 160 subjects.22 JMP v3.2 (SAS Institute Inc, Cary, NC) was used for
Of 29 participating hemodialysis facilities, 23 (79%) were free standing
(vs hospital based) and 21 (72%) were for profit. Of 53 participating nephrologists,
83% were men, 55% were white, and their mean age was 48 years.
Figure 1 illustrates the flow
of participants through the trial. A total of 169 patients completed the trial.
Eighty eligible patients did not complete the trial, either because they declined
to participate, were unable to participate because they were mentally impaired
or did not speak English, or died or moved prior to reaching the final evaluation
phase. The 80 nonparticipants were older than the 169 participants (65 years
vs 55 years, P<.001), but did not differ in sex,
race, cause of renal failure, years receiving dialysis, or baseline Kt/V.
Nine of the 53 nephrologists did not have any eligible patients.
Intervention and control patients had similar demographic and medical
characteristics, baseline Kt/V and facility Kt/V goals, and specific barriers
to adequate hemodialysis (Table 1).
The most common barrier was low prescription, present in 68% of intervention
patients and 75% of control patients (P = .33).
As seen in Table 2, intervention
patients had 2-fold larger increases in Kt/V compared with control patients
(+0.20 vs +0.10; 95% confidence interval for difference, 0.05-0.15; P<.001). Intervention patients were also more likely
to achieve their facility Kt/V goal (62% vs 42%, P
Among control patients, large body size was associated with a lower
likelihood of achieving the facility Kt/V goal. Sixty-two percent of control
patients in the lowest tertile of body weight achieved the facility Kt/V goal
compared with 31% of patients in both the middle and highest tertile (P = .04). Among intervention patients, there was no relationship
between body size and likelihood of achieving the facility goal (61%, 69%,
and 57% of patients in the lowest, middle, and highest tertile, respectively,
achieved the facility goal, P = .64).
Among the 121 subjects with low prescribed Kt/V at baseline (Table 2), intervention patients had nearly
3-fold larger increases in prescribed dialysis dose compared with control
patients (+0.16 vs +0.06, P<.001). Among the 57
subjects with catheters at baseline, intervention patients were 4 times more
likely to have catheters changed to fistulas or grafts compared with control
patients (28% vs 7%, P = .04). Among the 39 subjects
with shortened treatment time at baseline, both intervention and control patients
had similar increases in treatment time (+13 minutes vs +13 minutes, P = .94).
The number of barriers remaining at the end of the trial correlated
strongly with achieving the facility Kt/V goal among both intervention and
control patients (Figure 2). Eighty-two
percent of patients with no barriers remaining achieved the Kt/V goal, while
only 13% of patients with 3 barriers remaining achieved the Kt/V goal.
Even though our intervention was an overall success, specific barriers
persisted among some intervention patients. The reasons for failing to overcome
barriers are listed in Table 3.
The 4 possible ways to increase dialysis prescription are listed separately.
Twenty intervention patients did not receive higher-efficiency dialyzers.
In 1 case, the patient had a history of an allergic reaction to the higher-efficiency
dialyzer; in 4 cases, the nephrologist refused to prescribe a higher-efficiency
dialyzer; and in 15 cases, the facility did not use higher-efficiency dialyzers.
Ten patients did not have blood flow increased, due to medical limitations
(eg, poor function of fistula or graft) or patient refusal. Five patients
did not have dialysate flow increased because their facility used older equipment
that could not accommodate higher flows. Twenty-two patients did not have
treatment time increased because of patient refusal, physician refusal, or
a tight facility schedule that prevented increased treatment time.
Twenty-one patients did not have catheters changed to fistulas or grafts.
In 11 cases, patients had medical limitations such as lack of vascular sites
or frequent clotting; in 6 cases patients refused; and in 4 cases, the physicians
Fourteen patients continued to shorten treatment time. In 5 cases, patients
had medical limitations (eg, back pain from prolonged sitting) and in 9 cases,
patients refused to decrease shortening of treatment time.
There was no difference between intervention and control patients in
baseline or final assessments of any of the 7 quality-of-life subscales examined
(results not shown). In addition, there was no relationship between change
in treatment time and items related to burden of kidney disease or specific
dialysis-associated symptoms including cramping, nausea/vomiting, dizziness,
and feeling washed out.
This is the first randomized controlled trial of an intervention that
targets specific barriers to adequate hemodialysis. We demonstrated that addressing
3 common and easily identifiable patient-specific barriers resulted in 2-fold
larger increases in dialysis dose compared with usual care. Only 18% of patients
who overcame all their barriers still had inadequate dialysis doses (Figure 2). This suggests that other, as-yet
unidentified barriers are unlikely to be important impediments to adequate
dialysis. The finding of statistically significant and clinically important
effects despite a modest sample size and relatively brief 6-month follow-up
provides further evidence of the intervention's effectiveness. By enrolling
subjects receiving the lowest dialysis doses, we improved adequacy of dialysis
among patients at the highest risk of mortality, morbidity, and increased
health care costs.3,8 While some
patients may view dialysis treatment time as burdensome, we found that increasing
dialysis dose did not adversely affect patient quality of life.
By engaging the participation of virtually all dialysis facilities and
nephrologists in a large geographic area, we enhanced the generalizability
of our findings. With the exception of sex and race (Table 1), patient characteristics and facility characteristics are
comparable to national data.2 As expected,
men and black patients were overrepresented among those receiving inadequate
hemodialysis.25 This appears to be due to a
larger body size and more frequent shortening of treatment time by men and
black patients.16,26- 28
Despite a larger urea distribution volume, large patients in the intervention
group improved as much as smaller patients. By contrast, large patients did
not do as well as smaller patients in the control group.
Our results have important implications for patients, providers, and
health policy makers. Patients can improve dialysis adequacy by allowing increased
prescriptions, expediting conversion of catheters to fistulas or grafts when
possible, and minimizing shortening of treatments. Because symptoms such as
cramping tend to occur near the end of a dialysis treatment, some patients
are reluctant to increase treatment time for fear of experiencing more symptoms.6,29 Such patients may be reassured to
know that a longer treatment time was not associated with an increase in symptoms
or in perceived burden of kidney disease among study participants. Providers
can improve dialysis adequacy by routinely monitoring not only Kt/V but also
the 3 specific barriers to adequate dialysis. In particular, we found that
nephrologists were unaware of low prescriptions because they do not explicitly
calculate prescribed Kt/V as we did for this trial. In addition, it is imperative
to identify and eliminate provider-related impediments such as a lack of higher-efficiency
dialyzers or use of old machines. Policy makers can improve dialysis adequacy
by reimbursing facilities more for longer treatments and for using higher-efficiency
dialyzers (as opposed to the current fixed payment per treatment). Because
inadequate dialysis is independently associated with more frequent hospitalizations,
an increased reimbursement to outpatient dialysis facilities may be offset
by savings in inpatient expenditures. In a previous investigation, we estimated
that a 0.10 increase in Kt/V is independently associated with an $1880 per
patient decrease in annual inpatient expenditures.3
Thus, applying our intervention to all 33 000 patients receiving inadequate
dialysis may result in a $62 million savings in inpatient expenditures.
The randomized trial design allows us to differentiate the effect of
our intervention from the usual care provided by nephrologists. Inadequate
hemodialysis has been the focus of intense scrutiny and quality improvement
efforts at the national, regional, and facility levels for the last several
years.1,7 The improvements in
Kt/V among control patients suggest that these efforts may indeed help to
improve dialysis dose, although not as effectively as our targeted approach.30,31 Recruitment of both intervention
and control patients from the same geographic area further diminishes the
possibility that some external influence differentially affected the 2 groups.
The favorable change in both the end point (Kt/V) and the intermediate variables
(patient-specific barriers) strengthens the causal link between our intervention
and increased dialysis dose. Two other possible contributors to the observed
improvement in the control group are worth mentioning. First, control patients
may have improved in part because they and their physicians were being observed
(the Hawthorne effect).32 Second, it is possible
that control patients were influenced by intervention patients if they were
at the same facility (contamination). However, both of these effects would
tend to decrease the difference between control and intervention patients,
and therefore the measured effect size of 0.10 may underestimate the value
of our intervention.
While our intervention was an overall success, many intervention patients
failed to overcome specific barriers (Table
3). In addition, our intervention was no more effective than usual
care for addressing the shortened treatment time barrier. Further refinements
of our approach may be needed to increase its potency.
In conclusion, we recommend that nephrologists, individual hemodialysis
facilities, dialysis chains, and regulatory agencies monitor and address 3
specific barriers to adequate dialysis: underprescription, catheter use, and
shortening of treatment time. Although our modest sample size and brief follow-up
period are insufficient to demonstrate an impact on patient mortality and
morbidity, several large observational studies have demonstrated a link between
Kt/V and these patient outcomes.3,5,8,9
Thus, overcoming patient-specific barriers has the potential to enhance survival
and decrease both hospitalizations and inpatient expenditures. Our approach
of first identifying promising barriers in observational studies and then
testing interventions to overcome these barriers in randomized controlled
trials may also be applicable to quality improvement efforts in other medical