Content uploaded by Rae Seitz
Author content
All content in this area was uploaded by Rae Seitz on Apr 23, 2018
Content may be subject to copyright.
Increased Satisfaction with Care and Lower Costs: Results of a
Randomized Trial of In-Home Palliative Care
Richard Brumley, MD,
Susan Enguidanos, PhD, MPH,
w
Paula Jamison, BA,
w
Rae Seitz, MD,
z
Nora Morgenstern, MD,
§
Sherry Saito, MD,
z
Jan McIlwane, MSW,
§
Kristine Hillary, RNP,
and
Jorge Gonzalez, BA
w
OBJECTIVES: To determine whether an in-home pallia-
tive care intervention for terminally ill patients can improve
patient satisfaction, reduce medical care costs, and increase
the proportion of patients dying at home.
DESIGN: A randomized, controlled trial.
SETTING: Two health maintenance organizations in two
states.
PARTICIPANTS: Homebound, terminally ill patients
(N 5298) with a prognosis of approximately 1 year or less
to live plus one or more hospital or emergency department
visits in the previous 12 months.
INTERVENTION: Usual versus in-home palliative care
plus usual care delivered by an interdisciplinary team pro-
viding pain and symptom relief, patient and family educa-
tion and training, and an array of medical and social
support services.
MEASUREMENTS: Measured outcomes were satisfac-
tion with care, use of medical services, site of death, and
costs of care.
RESULTS: Patients randomized to in-home palliative care
reported greater improvement in satisfaction with care at
30 and 90 days after enrollment (Po.05) and were more
likely to die at home than those receiving usual care
(Po.001). In addition, in-home palliative care subjects were
less likely to visit the emergency department (P5.01) or be
admitted to the hospital than those receiving usual care
(Po.001), resulting in significantly lower costs of care for
intervention patients (P5.03).
CONCLUSION: In-home palliative care significantly in-
creased patient satisfaction while reducing use of medical
services and costs of medical care at the end of life. This
study, although modest in scope, presents strong evidence
for reforming end-of-life care. J Am Geriatr Soc 55:
993–1000, 2007.
Key words: palliative care; in-home services; patient sat-
isfaction; end-of-life care
It has been widely recognized that our current medical
care structure is inadequate in meeting the needs of ter-
minally ill patients and reducing the cost of care at the end
of life.
1
Despite the existence of hospice as a Medicare
benefit for nearly 2 decades, the program remains under-
used. Approximately 60% of all deaths occur in the hos-
pital,
2
yet most patients express a preference to die at
home.
3–6
Although hospice programs aim to provide pal-
liative services in the last 6 months of life, the median length
of stay in the program is 22 days, and 35% of patients die
within the first 7 days after hospice admission.
7
Hospice
patients with a short length of stay often require intensive
care to initiate the care plan, resulting in higher per diem
costs of care than for patients who receive longer periods of
stabilized, low-cost palliative care.
8,9
The low enrollment in
hospice services and the short length of time enrolled before
death attest to the need for end-of-life care programs that
address these access barriers. In addition, recent studies
have found that more end-of-life programs are needed to
provide alternatives to hospice that do not require forgoing
life-sustaining treatment.
10
Although several studies have reported that end-of-life
care programs improve patient outcomes, these studies have
significant methodological weaknesses, limiting their gen-
eralizability.
11
Specifically, there has been a noticeable lack
of comprehensive empirical evidence confirming the clinical
benefits and demonstrating the cost effectiveness of these
models of care. The absence of rigorous research evaluating
the effectiveness of these programs has restricted the ability
and motivation of healthcare providers to replicate and
adopt these models as standard practice.
The purpose of this study was to test an in-home
palliative care model at two sites using a randomized,
Address correspondence to Susan Enguidanos, PhD, Partners in Care Foun-
dation, Director, Research Center, 732 Mott Street, Suite 150, San Fernando,
CA 91340. E-mail: Senguidanos@picf.org
DOI: 10.1111/j.1532-5415.2007.01234.x
Kaiser Permanente Southern California Medical Group, Downey, Califor-
nia;
w
Partners in Care Foundation, San Fernando, California;
z
Kaiser Per-
manente Hawaii Medical Group, Honolulu, Hawaii;
§
Kaiser Permanente
Colorado Medical Group, Aurora, Colorado.
JAGS 55:993–1000, 2007
r2007, Copyright the Authors
Journal compilation r2007, The American Geriatrics Society 0002-8614/07/$15.00
controlled design. Standard care was compared with stan-
dard care plus an in-home palliative care program to de-
termine the program’s ability to improve patient outcomes
and reduce the costs of medical care at the end of life. Spe-
cifically, it was hypothesized that the palliative care pro-
gram would increase patient satisfaction, reduce costs of
medical care, and increase the proportion of terminally ill
patients dying at home.
METHODS
This was a randomized, controlled trial conducted at two
separate managed care sites to test the replicability and the
effectiveness of an In-home Palliative Care (IHPC) pro-
gram. Specific hypotheses tested for this study were that
Late-stage patients with chronic obstructive pulmonary
disease (COPD), congestive heart failure (CHF), or can-
cer enrolled in the IHPC program will experience higher
satisfaction with care than patients receiving usual and
customary care (usual care).
Medical costs will be lower for end-of-life patients en-
rolled in the IHPC program than for end-of-life patients
receiving usual care.
The settings for this trial were two group-model, closed-
panel, non-profit health maintenance organizations
(HMOs) providing integrated healthcare services in Ha-
waii and Colorado. The Colorado site has more than 500
physicians representing all medical specialties and subspe-
cialties in 16 separate ambulatory medical offices spread
across a greater metropolitan area. The HMO contracts
with outside providers for emergency department, hospital,
home health, and hospice care to serve its 477,000-person
membership, which spans the six-county Denver metropol-
itan area.
The Hawaii site is located on Oahu and serves approx-
imately 224,000 members, with 12 medical offices in Oahu,
three in Maui, and three on the Big Island. Three hundred
seventeen medical group physicians provide care. In con-
trast to Colorado, the HMO provides all outpatient and
most inpatient care. The Hawaii site operates a 217-bed
medical center and also has an internal home health agency,
accepting referrals from hospital- and clinic-based medical
group physicians. Referrals are processed via a referral
center staffed by home care nurses. The Hawaii site does
not have an internalized hospice agency and refers patients
to outside providers for hospice care only.
Principal investigators in Southern California, where
data collection and analysis were conducted, coordinated
the study. The institutional review boards of the HMO in
Hawaii, Colorado, and the external evaluator approved this
study. Participants were enrolled and followed from Sep-
tember 2002 to August 2004. Patients eligible to participate
in the study must have had a primary diagnosis of CHF,
COPD, or cancer and a life expectancy of 12 months or less,
have visited the emergency department or hospital at least
once within the previous year; and scored 70% or less on
the Palliative Performance Scale.
12,13
The Palliative Perfor-
mance Scale is a modified Karnofsky scale that ranks the
patient’s health condition from 0 (death) to 100 (normal).
This scale was used to assess the patients’ severity of illness.
To determine life expectancy, the primary care physician
was asked, ‘‘Would you be surprised if this patient died in
the next year?’’ This indicator has been used widely to as-
certain appropriateness for end-of-life care.
14,15
Patients
with physician responses indicating no surprise if the pa-
tient died within the next year were included in the study.
Discharge planners, primary care physicians, and other
specialty physicians referred potentially eligible terminally
ill patients to the study. All patients referred were assessed
for study eligibility. For those meeting the initial criteria, the
intake clerk contacted the primary care physician to deter-
mine the prognosis. Once eligibility was determined, the
intake clerk gained informed consent from the patient to
participate in the study. The intake clerk then contacted
evaluators, who randomly assigned patients to the pallia-
tive care intervention or usual care. Group assignment was
determined by blocked randomization using a computer-
generated random number chart, stratified according to
study site. Based on methods established previously,
16
it
was determined that, using a significance criterion of .05, a
sample size of 300 would be necessary for a statistical
power of 0.80, using nondirectional (two-tailed) tests to
detect whether the intervention had a significant effect on
medical care costs.
Intervention
Each patient enrolled in the intervention arm received cus-
tomary and usual standard care within individual health
benefit limits in addition to the IHPC program. The IHPC
program is an interdisciplinary home-based healthcare pro-
gram designed to provide treatment with the primary intent
of enhancing comfort, managing symptoms, and improving
the quality of a patient’s life. Modeled after hospice pro-
grams in that it offers pain management and other comfort
care in the patient’s home, the IHPC program features three
important modifications, all of them intended to increase
access and timely referrals to the program.
Physicians are not required to give a 6-month prognosis.
Recognizing that it is often difficult to estimate life ex-
pectancy, referral guidelines were expanded to target
patients earlier in their disease process, with an esti-
mated 12-month life expectancy.
Although the IHPC program emphasizes much-
improved pain control and other symptom manage-
ment, patients do not have to forgo curative care, as they
do in hospice programs.
Patients are assigned a palliative care physician who
coordinates care from a variety of healthcare providers,
including specialists and the patients’ primary care phy-
sician, thus preventing the service fragmentation that
often occurs in healthcare systems. The structure of the
IHPC program allows patients to maintain their prima-
ry care provider while also receiving home visits from
the palliative care physician.
The IHPC program uses an interdisciplinary team ap-
proach, with the core care team consisting of the patient
and family plus a physician, nurse, and social worker with
expertise in symptom management and biopsychosocial in-
tervention. The core team is responsible for coordinating
and managing care across all settings and providing assess-
ment, evaluation, planning, care delivery, follow-up,
994 BRUMLEY ET AL. JULY 2007–VOL. 55, NO. 7 JAGS
monitoring, and continuous reassessment of care. Compre-
hensive education and discussions focus on identifying
goals of care and the expected course of the disease and
expected outcomes, as well as the likelihood of success of
various treatments and interventions.
Upon admission, the team assesses the physical, med-
ical, psychological, social, and spiritual needs of the patient
and family. All patients received initial assessments from
physicians, nurses, and social workers. Additional team
members, including spiritual counselor or chaplain,
bereavement coordinator, home health aide, pharmacist,
dietitian, volunteer, physical therapist, occupational thera-
pist, and speech therapist, join the core care team in service
provision as needed. The team convenes to develop a care
plan in accordance with the wishes of the patient and the
family. Frequency of subsequent medical visits is based on
the individual needs of the patient. Physicians conduct
home visits and are available along with nursing services on
a 24-hour on-call basis. In addition, advanced care planning
is provided that involves patients and their families in mak-
ing informed decisions and choices about care goals and
end-of-life care.
The team provides education, support, and medical
care to the patients and their families. Additionally, patients
and families are trained in the use of medications, self-
management skills, and crisis intervention in the home with
the goal of stabilizing the patient and minimizing excessive
emergency department visits and acute care admissions.
Participants enrolled in the IHPC arm received palliative
care until death or transfer to a hospice program. (For more
information on this model, see
17,18
.)
Usual Care
Usual care consisted of standard care to meet the needs of
the patients and followed Medicare guidelines for home
healthcare criteria. These services included various amounts
and levels of home health services, acute care services, pri-
mary care services, and hospice care. Patients were treated
for conditions and symptoms when they presented them to
attending physicians. Additionally, they received ongoing
home care when they met the Medicare-certified criteria for
an acute condition.
Data Collection
Data were collected from patient interviews and from the
HMO service utilization databases at each site. Interviews
were conducted via telephone within 48 hours of study
enrollment and every 30, 60, 90, and 120 days to gather
demographic information and satisfaction with services.
Undergraduate- and graduate-level research assistants,
blinded to group assignments, were recruited and trained
to conduct telephone interviews with the patient or, if the
patient was unable to participate, the primary caregiver.
Interviews were approximately 20 minutes long. Site of
death was obtained from HMO records, death certificates,
and family report.
Utilization Data
Service utilization data for each subject were collected ret-
rospectively from the HMO mainframe database. Medical
service use data were collected from the time the patient
enrolled in the study until the time of death or the end of the
study period. Medical service use data included costs for all
standard medical care as well as the costs associated with
the palliative care program. Service data included number
of emergency department visits, physician office visits, hos-
pital days, skilled nursing facility days, home health and
palliative visits, palliative physician home visits, and days
on hospice. Service costs were calculated using actual costs
for contracted medical services (services provided by non-
HMO contracted facilities in Colorado) and proxy cost es-
timates for all services provided within the HMO. Because
services provided within the HMO are not billed separately,
it was necessary to use proxy costs. Costs were based on
figures from 2002. Hospitalization and emergency depart-
ment cost estimates were calculated using aggregated data
from more than 500,000 HMO patient records and include
ancillary services such as laboratory and radiology. Costs of
physician office visits included nurse and clerk expenses.
Home health and palliative care visits were calculated using
average time spent on each visit and multiplying that by the
cost for each discipline’s reimbursement rate. Proxy costs
generated for hospital days and emergency department vis-
its were significantly lower than the actual costs received
from contracted providers. A total cost variable was con-
structed by aggregating costs for physician visits, emergen-
cy department visits, hospital days, skilled nursing facility
days, and home health or palliative days accumulated from
the point of study enrollment until the end of the study
period or death.
Enrollment in hospice was gathered retrospectively and
was only available from one study site because of difficulty
obtaining data from community hospice providers at the
second site.
Satisfaction Instrument
The Reid-Gundlach Satisfaction with Services instrument
19
was used to measure study group member satisfaction, rat-
ing overall satisfaction with services, perception of service
providers, and likelihood of positive recommendations of
services to others. This instrument has been employed in
previous studies examining satisfaction with end-of-life
care programs.
17,20,21
Satisfaction with care was calculated
by adding the score of 12 of the 13 items (one item was a
qualitative measure) on the instrument for a total possible
score of 48. This method has been employed in previous
studies using this instrument.
Severity of Illness
The Palliative Performance Scale,
12,13
described earlier in
this manuscript, was used to measure severity of illness.
Statistical Analysis
Differences between study group sample characteristics
were analyzed using two-tailed ttests for continuous vari-
ables where the distribution was normal. Chi-square tests
were used to determine significant differences between dis-
crete variables. Satisfaction with services was analyzed in
two ways. First, to determine the clinical meaningfulness of
the change in satisfaction according to study group, satis-
faction scores were dichotomized as recommended previ-
ously
22
and used in other studies,
23
with those with a total
score of 37 or above categorized as very satisfied. Next,
RANDOMIZED IN-HOME PALLIATIVE CARE TRIAL 995JAGS JULY 2007–VOL. 55, NO. 7
logistic regression models were developed for dichotomous
satisfaction variables using baseline and 30-, 60-, and 90-
day follow-up measures. Linear regression was employed to
determine the effect of study group on the number of hos-
pital inpatient days and emergency department days while
controlling for length of time enrolled in the study (survival)
and demographic variables. Logistical regression was used
to determine study group likelihood of dying at home.
Kaplan-Meier survival analysis, using the log rank statistic,
tested for study group differences in survival time. All study
participants were included in the analysis, with those who
survived to the end of the study period censored on the last
day of the study.
Medical cost data are generally skewed, requiring trans-
formation of the dependent variable for better data fit. As
expected, analysis of total medical costs indicated that they
were extremely right-skewed; hence, a log transformation of
medical costs and Duan’s smearing estimate followed linear
regression to determine whether transformation of data re-
sulted in a better fit. A comparison of models revealed that,
although the log transformation resulted in a normal distri-
bution, the ordinary least squares regression explained a
larger amount of the variance (higher coefficient of determi-
nation; R
2
) and had lower error. Further comparison of the
residual distributions (the difference between the actual val-
ue of the dependent variable and its value as estimated by the
equation) revealed a linear relationship between the non-
transformed costs. Based on these findings, ordinary least
squares regression was used in the final analysis, a decision
supported by other studies.
24
Owing to the differences in
survival time between study groups, analysis of service use
and medical care costs were adjusted for length of time on
the program before death or the end of the study period.
Analyses were conducted using SPSS 10.1 statistical software
package (SPSS Inc., Chicago, IL) and LIMDEP, version 8.0
(Econometric Software, Inc., Plainview, NY).
RESULTS
From September 2002 through March 2004, 718 patients
were referred to the study. Of these, 408 were excluded; 196
did not meet study eligibility criteria, 67 were eligible for
and admitted to hospice care, 59 refused, 38 died before
enrollment, 26 were part of another research project ter-
minating their eligibility for participation in this study, and
22 moved out of the area or could not be contacted. As a
result, 310 terminally ill participants were randomly as-
signed to usual care (n 5155) or the in-home palliative care
program (n 5155). Of these, eight intervention group
members died before receiving any palliative care, and five
usual care members withdrew from the study, leaving 297
available for analysis (Figure 1).
The sample consisted of almost equal numbers of men
(51%) and women (49%), with a mean age standard de-
viation of 74 12.0. Thirty-seven percent belonged to an
ethnic minority group; 18% were Asian/Pacific Islanders,
13% Hawaiian, 4% Latino, and 2% other. Fifty-two per-
cent were married, 29% widowed, and 15% single or di-
vorced. Forty-seven percent were referred to the program
with a diagnosis of cancer, 33% with CHF, and 21% with
COPD. Seventy-six percent lived in their own home or
apartment, and 8% lived in the home of a family member;
74% resided with a family member, primarily a spouse or a
child, and 26% lived alone. Thirty-three percent reported
having an annual income of $20,000 or less. Participants
came from an array of educational backgrounds; approx-
imately 22% did not complete high school, 41% were high
school graduates, and 36% had some college or postgrad-
uate education (Table 1).
Overall, the majority of the sample demographics were
consistent at both study sites, although there was some
variation. Sixty-three percent of participants in Hawaii
were minorities, compared with 10% in Colorado. This
difference in ethnic distribution is reflective of the larger
demographics within each of the states. Eighty-two percent
of participants in Hawaii were living with a family member,
compared with 72% in Colorado. Finally, 27% of partic-
ipants in Colorado suffered from COPD, versus 15% of
participants in Hawaii. There were no significant differ-
ences between sites in age, education, marital status, in-
come, or housing status (Table 1).
During the course of the study, 75% (n 5225) of the
participants died. There were no significant differences be-
tween study groups in terms of the portion of patients dying
during the study period, although differential survival pe-
riods after enrollment in the study were found using inde-
pendent-sample ttests, with those enrolled in the
intervention surviving an average of 196 164 days and
those in usual care surviving an average of 242 200 days
(P5.03). Results of the Kaplan-Meier survival analysis did
not show significant differences in survival time between
study groups (log rank test 52.98; P5.08), although sub-
sequent analysis controlled for survival days due to the
strong trend toward differences and its potential effect on
use of medical services and costs of medical care.
Excluded
(n = 408)
Not meeting inclusion criteria
(n = 196)
Refused to participate
(n = 59)
Other reasons*
(n = 153)
Assigned to intervention
(n = 155)
Withdrew
(n = 2)
Died before intervention
(n = 8)
Assigned to usual care
(n = 155)
Withdrew from study
(n = 3)
Analyzed (n = 152)
Analyzed (n = 145)
Assessed for eligibility
(n = 718)
Randomized
(n = 310)
Figure 1. Patient flowchart.
Other reasons includes those re-
ferred to hospice (n 567), those who died (n 538), those who
were already in another study (n 526), and those who moved
out of the area or could not be contacted (n 522).
996 BRUMLEY ET AL. JULY 2007–VOL. 55, NO. 7 JAGS
Baseline Measures
There were no significant differences between study groups
in baseline measures other than satisfaction. Satisfaction
with services was measured at baseline after study assign-
ment. Those randomized to intervention demonstrated sig-
nificantly higher satisfaction with services at baseline than
those assigned to usual care (P5.03). Member awareness of
the results of randomization may have influenced the higher
level of satisfaction at baseline in those in the palliative care
group.
Satisfaction with Care
Analysis of satisfaction data included satisfaction at base-
line (n5277) and 30 days (n 5216), 60 days (n5168), and
90 days (n5149) after study enrollment. Significant reduc-
tion in sample size at 120 days (n5136) resulted in the
exclusion of this data in analyses. There was no significant
difference in the portion of participants according to study
group reporting to be very satisfied at baseline or at 60 days
after enrollment (odds ratio (OR)51.79; 95% confidence
interval (CI)50.65–4.96; P5.26), although rates of satis-
faction increased in the intervention group at 30 days
(OR53.37, 95% CI 51.42–8.10; P5.006) and 90 days
(OR53.37, 95% CI 50.65–4.96; P5.03) after enrollment,
with 93% of those enrolled in the palliative care group very
satisfied with care at 90 days after enrollment, compared
with 81% of usual care patients (Figure 2).
Service Use
Bivariate analysis revealed significant differences between
groups in terms of service use. Twenty percent of palliative
care members went to the emergency department, com-
pared with 33% of usual care members (P5.01; Cramer’s
V5.15). Similarly, 36% of those receiving palliative care
were hospitalized, compared with 59% of those enrolled in
usual care (Po.001; Cramer’s V5.23). Number of days in
the study was significantly different according to study
group, as well. Those enrolled in the IHPC group remained
in the study for 196 days on average, whereas those in the
usual care group were in the study an average of 242 days.
Because of these differences, additional analysis was con-
ducted after controlling for length of time on the program
Table 1. Baseline Characteristics
Characteristic
Colorado
(n 5147)
Hawaii
(n 5150)
Usual Care
(n 5152)
Intervention
(n 5145)
Total
(N 5297)
Female, n (%) 71 (48) 74 (49) 81 (53) 65 (45) 146 (49)
Age, mean SD 74.1 10.8 74.3 13.1 73.7 13.0 73.9 11.1 73.8 12.1
Racial minority, n (%) 14 (10)
95 (63)
53 (35) 56 (39) 109 (37)
Married, n (%) 83 (58) 72 (51) 73 (48) 82 (57) 155 (52)
Primary diagnosis, n (%)
Cancer 65 (44) 73 (48) 74 (49) 64 (44) 138 (47)
Congestive heart failure 42 (29) 55 (37) 52 (34) 45 (31) 97 (33)
Chronic obstructive pulmonary disease 40 (27)
22 (15)
26 (17) 36 (25) 62 (21)
Education, mean SD 12.0 2.1 11.6 2.5 11.9 2.2 11.8 2.4 11.9 2.3
Lives with family member, n (%) 103 (72)
116 (82)
105 (69) 114 (79) 219 (74)
Lives in own house/apartment, n (%) 113 (77) 114 (76) 113 (74) 114 (79) 227 (76)
Annual income o$20,000, n (%) 46 (45) 53 (45) 53 (35) 46 (32) 99 (33)
Palliative Performance Scale score, mean SD 62.5 11.5 54.2 11.9 58.5 12.0 57.8 13.1 58.2 12.5
Satisfaction, mean SD 39.4 6.0 40.9 5.3 39.3 6.2 40.8 5.2
40.1 5.7
Po.05.
SD 5standard deviation.
80.4 74.1
93.1
80
92.3
87
93.4
80.8
0
10
20
30
40
50
60
70
80
90
100
Palliative Care Usual Care
Percent Satisfied
Baseline 30 Days 60 Days 90 Days
Figure 2. Percentage very satisfied at enrollment (n 5277), 30 days (n 5216), 60 days (n 5168), and 90 days postenrollment
(n 5149) according to study group.
RANDOMIZED IN-HOME PALLIATIVE CARE TRIAL 997JAGS JULY 2007–VOL. 55, NO. 7
(survival), age, and severity of illness. Linear regression re-
vealed that enrollment in the IHPC reduced hospital days by
4.36 (Po.001; R
2
50.14) and emergency department visits
by 0.35 (P5.02; R
2
50.04) after adjusting for survival,
age, and severity of illness.
Although enrollment in hospice was not a specific aim
of this project, rates of enrollment were reviewed. Hospice
data were available from one of the study sites, with anal-
ysis of differences between the portions enrolled in hospice
(25% of intervention vs 36% of usual care, P5.15) and
days in hospice (t5.52; P5.60) before death revealing no
significant differences between study groups.
Costs of Care
Significant differences between palliative and usual care
members in cost of care (t53.63, Po.001) were noted.
Owing to differences in time enrolled in the study, a linear
regression was conducted to determine the portion of costs
explained by study group, controlling for days on service
(survival), age, severity of illness (measured using the Pal-
liative Performance Scale), and primary disease. Days en-
rolled was significantly correlated with age and severity of
illness, although the associations between these variables
were weak (r50.22 and 0.20, respectively). Three vari-
ables were significant in the regression model: age (although
the effect size was small), days enrolled (survival), and en-
rollment in the IHPC program (Table 2). This analysis re-
vealed that overall costs of care for those enrolled in the
IHPC program were 33% less than those receiving standard
care (P5.03; 95% CI 5$12,411 to $780; R
2
50.16).
The adjusted mean cost for patients enrolled in the pallia-
tive care group was $12,670 $12,523, compared with
$20,222 $30,026 for usual care. Figure 3 represents the
average cost of care per member per day according to study
group. The average cost per day incurred by palliative care
recipients ($95.30) was significantly lower than that of
usual care group members ($212.80) (t52.417; P5.02).
Site of Death
During the course of the study, 75% of study members in-
cluded in the final analysis died. For 98% of these persons,
site of death data were available. Seventy-one percent of
IHPC participants died at home, compared with 51% of
those receiving usual care (Po.001). Bivariate logistic anal-
ysis confirmed that patients enrolled in the IHPC program
were significantly more likely to die at home and less likely
to die in an acute care facility. Furthermore, after control-
ling for age, survival time, and medical conditions, IHPC
participants were 2.2 times as likely to die at home as those
receiving usual care (OR 52.20, 95% CI 51.3–3.7;
R
2
50.27, Po.001).
DISCUSSION
This study examining the effect of the IHPC revealed sev-
eral positive findings. The IHPC intervention improved pa-
tient satisfaction at 30 and 90 days after enrollment,
improved the likelihood of dying at home, and significantly
reduced the cost of care overall and by average cost per day.
First, providing an interdisciplinary palliative care team
within the home of terminally ill homebound patients ear-
lier in the disease trajectory has a positive effect on patient
satisfaction with medical care in addition to influencing
costs of care at the end of life. Recent studies have found
that, although costs of care vary from state to state and
from hospital to hospital, they remain high in the last 2
years of life.
25
In addition, previous studies focusing on
costs of care in the last year of life found that average per-
member costs have remained constant over the past decade,
representing approximately 25% of all Medicare expendi-
tures.
26
Moreover, a recent study examining costs of care in
the last 2 years of life estimated average costs to be ap-
proximately $58,000.
25
This finding suggests that end-of-
life care programs should not be limited to the last 6 months
of survival, because costs associated with end-of-life care
are likely to accrue over the last 2 years of life. This study
supports findings from a previous study of this model
17
that
found significantly lower costs for palliative care than for
standard care.
There was a strong trend toward shorter survival for
those in the palliative care group (196 days vs 242 days)
after study enrollment. The differential in survival period
after study enrollment may be attributed to several factors.
Table 2. Predictors of Medical Costs After Study Enrollment (N 5297)
Variable Mean Cost 95% Confidence Interval P-Value R
2
Age 312 (547 to 761.02) .01 .002
Days enrolled 42 (27–57) .001 .125
Health status 2,323 (5,524–878) .15 .130
Congestive heart failure (vs cancer) 5,255 (1,297–11,808) .12 .138
Chronic obstructive pulmonary disease (vs cancer) 3,294 (3,648–10,237) .35 .138
Palliative (vs usual care) 7,552 (12,730 to 2,374) .004 .160
Note: Dependent variable, total medical costs.
Mean costs are measured in dollars.
$95.30
$212.80
$0.00
$20.00
$40.00
$60.00
$80.00
$100.00
$120.00
$140.00
$160.00
$180.00
$200.00
$220.00
$240.00
Mean Cost per Day
Palliative Care
Usual Care
Figure 3. Average adjusted cost of care per day according to
study group (n 5297).
998 BRUMLEY ET AL. JULY 2007–VOL. 55, NO. 7 JAGS
First, although the Palliative Performance Scale was used to
measure severity of illness, the two groups may have had
several differences in disease severity and range of medical
conditions that were not collected as part of this study. In
addition, the data do not take into account personal pref-
erences for care at the time of study enrollment and changes
in these preferences throughout the course of the study.
Moreover, the palliative care intervention contains strong
components, such as patient education, ongoing conversa-
tions about care preferences, and care plans, that are de-
veloped to ensure adherence to patient preferences and
directives, all of which may influence patient survival time.
Several studies have found that older patients would choose
pain and symptom relief and comfort care over aggressive
treatment to extend life.
27,28
Thus, the intervention com-
ponents that focused on delineating and following patient
care preferences also may have affected survival time. This
notion was supported in a chart review of a sample of 90
study participants that found that IHPC patients had fewer
911 calls and fewer life-sustaining interventions conducted
in the emergency department or intensive care unit.
Although enrollment in hospice was not a stated goal of
this program, a retrospective examination of enrollment
rates in participants of one study site was conducted that
found no significant difference in the rate of hospice en-
rollment between groups. This may be attributed to the fact
that the focus was on implementing the in-home palliative
care program rather than facilitating transfer to hospice and
to the structure of the in-home palliative care program.
Enrollment in hospice would entail changing care providers
from those employed at the managed care organization to a
care team in a free-standing outside (non-managed care)
organization.
Satisfaction with care improved significantly for those
enrolled in the intervention arm of this study at 30 days and
remained high, with more than 90% reporting being very
satisfied with their medical care. At 30 and 90 days, indi-
viduals in the palliative study group were three times as
likely to report high levels of satisfaction. This may be at-
tributable to several factors. As suggested above, the pal-
liative care model itself may be more conducive to the needs
and preferences of the patients. One study
29
found that the
most important elements in end-of-life care identified by
seriously ill patients and their families related to trust in the
treating physician, avoidance of unwanted life support,
effective communication, continuity of care, and life com-
pletion, all of which are core components of the IHPC pro-
gram. Another positive finding from the current study was
the ability to enroll diverse participants and retain them in
the intervention arm. Satisfaction appeared to be high in
this diverse population.
This study was limited in several respects. The ade-
quate sample of individual minority groups limited further
analysis of ethnic variation. This area merits additional in-
vestigation and attention given the IHPC program’s appar-
ent success in enrolling a wide range of ethnic groups. This
study was conducted within closed-system managed care
settings; as a result, it may be less generalizable to all
healthcare settings, and the relative cost savings may not be
realized across other settings. Additionally, the use of proxy
costs of care calculated from aggregated patient records
further limits the ability to generalize findings across set-
tings. It also did not extend to an examination of the effect
of the individual model components (such as 24-hour call
center and physician home visits) on patient outcomes; this
element is a critical next step in determining what aspects of
the model are associated with key outcomes. Relying on
death at home as a measurement of patient preferences for
site of death also presented a limitation. Although a better
measurement would have been death in the patient’s pre-
ferred locale, these data were not collected as part of this
study. Finally, this study was limited in the lack of accurate
hospice data available at one of the sites.
This is one of the first rigorous studies to examine the
effectiveness of an in-home, community-based, palliative
care program for terminally ill individuals. It provides
strong clinical and financial evidence supporting the pro-
vision of palliative care in the home of terminally ill patients
with cancer, COPD, and CHF with a life expectancy of
approximately 1 year. It also suggests major policy impli-
cations for reforming end-of-life care. Evidence provided
here and in a previous study
17
supports the need for fun-
damental changes in the design of our healthcare system by
adjusting our current hospice benefit to better meet the
needs of patients or developing a new, ‘‘pre-hospice’’ pal-
liative care benefit that provides a bridge between standard
medical care and hospice care.
ACKNOWLEDGMENTS
We would like to acknowledge the many terminally ill pa-
tients and their families who participated in this study.
Their generosity in sharing their perspectives, medical in-
formation, and time was an invaluable contribution. In ad-
dition, we would like to thank George Shannon, PhD, for
his editorial comments.
Financial Disclosure: This study was funded by the
Kaiser Permanente Garfield Memorial Fund. Richard
Brumley, MD, Nora Morgenstern, MD, Sherry Saito,
MD, and Rae Seitz, MD, are employed as physician
partners in the Permanente Medical Group of Southern
California, Colorado, and Hawaii, respectively. Susan
Enguidanos, PhD, is employed by Partners in Care Foun-
dation and conducted this work through a subcontract with
the Garfield Memorial Fund. Paula Jamison, MA Candi-
date, and Jorge Gonzalez, BA, are employed by Partners in
Care Foundation and serve a consultative role through this
employment to Kaiser Permanente. Kristine Hillary, RNP, is
employed by Kaiser Permanente. Janet McIlwaine, MSW,
was employed by Kaiser Permanente.
Author Contributions: Richard Brumley was the prin-
cipal investigator on the study and was responsible for the
design and supervision of the study intervention. Susan En-
guidanos was the co-investigator on the project and was
responsible for the study design, overall study implemen-
tation, data collection, analysis, and preparation of the
manuscript. Paula Jamison was the project manager and
data coordinator on the study and assisted in the prepara-
tion of the manuscript. Nora Morgenstern, Sherry Saito,
Rae Seitz, and Jan McIlwaine were co-investigators and
were responsible for intervention implementation at the lo-
cal sites, including participant eligibility screening and en-
rollment in the study. Kristine Hillary provided clinical
training and oversight on the palliative care intervention.
RANDOMIZED IN-HOME PALLIATIVE CARE TRIAL 999JAGS JULY 2007–VOL. 55, NO. 7
Jorge Gonzalez supervised day-to-day data collection
among study participants and oversaw data entry and da-
ta preparation for analysis.
Sponsor’s Role: The sponsor assisted with research site
selection and provided fiscal support for the study.
REFERENCES
1. Robert Wood Johnson Foundation, Last Acts Coalition. Means to a Better
End: A Report on Dying in America Today, November 2002.
2. Weitzen S, Teno JM, Fennell M et al. Factors associated with site of death: A
national study of where people die. Med Care 2003;41:323–335.
3. Hays JC, Galanos AN, Palmer TA et al. Preference for place of death in a
continuing care retirement community. Gerontologist 2001;41:123–128.
4. Gallup GJ. Spiritual Beliefs and the Dying Process: A Report on a National
Survey. The Nathan Cummings Foundation and Fetzer Institute, 2000.
5. Karlsen S, Addington-Hall J. How do cancer patients who die at home differ
from those who die elsewhere? Palliat Med 1998;12:279–286.
6. Townsend J, Frank AO, Fermont D et al. Terminal cancer care and patients’
preference for place of death: A prospective study. BMJ 1990;301:415–417.
7. NHPCO. What Is Hospice and Palliative Care? Vol. 2004. National Hospice
and Palliative Care Organization, 2004.
8. Cheung L, Fitch K, Pyenson B. National Hospice and Palliative Care Orga-
nization. The costs of hospice care: An actuarial evaluation of the Medicare
hospice benefit, 2001.
9. Huskamp HA, Beeuwkes Buntin M, Wang V et al. Robert Wood Johnson
Foundation, The Commonwealth Fund. Providing care at the end of life: An
assessment of Medicare coverage and reimbursement methods, October 2000.
10. Casarett D, Van Ness PH, O’Leary JR et al. Are patient preferences for life-
sustaining treatment really a barrier to hospice enrollment for older adults
with serious illness? J Am Geriatr Soc 2006;54:472–478.
11. Morrison RS. Palliative care outcomes research: The next steps. J Palliat Med
2005;8:13–16.
12. Anderson F, Downing M. Palliative performance scale (PPS): A new tool.
J Palliat Care 1996;12:5–11.
13. Morita T, Tsunoda J, Inoue S et al. Contributing factors to physical symptoms
in terminally-ill cancer patients. J Pain Symptom Manage 1999;18:338–346.
14. Lynn J, Lynch Schuster J, Kabcenell A. Improving Care for the End of Life: A
Sourcebook for Health Care Managers and Clinicians. New York: Oxford
Press, 2000.
15. Weiche R, Mundy B, Skokan L et al. Identifying patients nearing the end of life
from congestive heart failure or chronic obstructive pulmonary disease. J Gen
Intern Med 2000;15:9–10.
16. Cohen J. Statistical Power Analysis for the Behavioral Sciences, 2nd Ed.
Hillsdale, NJ: Erlbaum, 1988.
17. Brumley RD, Enguidanos S, Cherin D. Effectiveness of a home-based palliative
care program for end-of-life. J Palliat Med 2003;6:715–724.
18. Cherin DA, Enguidanos SM, Jamison P. Physicians as medical center ‘‘Ex-
tenders’’ in end-of-life care: Physician home visits as the lynch pin in creating
an end-of-life care system. Home Health Care Serv Q 2003;23:41–53.
19. Reid PN, Gundlach JH. A scale for the measurement of consumer satisfaction
with social services. J Soc Serv Res 1983;7:37–54.
20. Cherin D, Enguidanos S, Brumley R. Reflection in action in caring for the
dying: Applying organizational learning theory to improve communications in
terminal care. Home Health Care Serv Q 2001;19:65–78.
21. Cherin DA, Huba GJ, Steinberg J et al. Satisfaction with services in innovative
managed care programs for groups of traditionally underserved individuals
with HIV/AIDS: Empirical models. Home Health Care Serv Q 2001;19:
103–125.
22. Schunemann HJ, Akl EA, Guyatt GH. Interpreting the results of patient re-
ported outcome measures in clinical trials: The clinician’s perspective. Health
Qual Life Outcomes 2006;4:62.
23. Unutzer J, Katon W, Callahan CM et al. Collaborative care management of
late-life depression in the primary care setting: A randomized controlled trial.
JAMA 2002;288:2836–2845.
24. Wagner TH, Chen S, Barnett PG. Using average cost methods to estimate
encounter-level costs for medical-surgical stays in the VA. Med Care Res Rev
2003;60(3 Suppl):15S–36S.
25. Wennberg JE, Fisher ES, Baker L et al. Evaluating the efficiency of California
providers in caring for patients with chronic illnesses. Health Aff (Millwood)
2005;Jul-Dec Suppl Web Exclusives:W5-526-43.
26. Lubitz JD, Riley GF. Trends in Medicare payments in the last year of life.
N Engl J Med 1993;328:1092–1096.
27. Rose JH, O’Toole EE, Dawson NV et al. Perspectives, preferences, care prac-
tices, and outcomes among older and middle-aged patients with late-stage
cancer. J Clin Oncol 2004;22:4907–4917.
28. Nahm ES, Resnick B. End-of-life treatment preferences among older adults.
Nurs Ethics 2001;8:533–543.
29. Heyland DK, Dodek P, Rocker G et al. What matters most in end-of-life care:
Perceptions of seriously ill patients and their family members. Can Med Assoc
J 2006;174:627–633.
1000 BRUMLEY ET AL. JULY 2007–VOL. 55, NO. 7 JAGS
