Mitchell SL, Teno JM, Roy J, Kabumoto G, Mor V. Clinical and Organizational Factors Associated With Feeding Tube Use Among Nursing Home Residents With Advanced Cognitive Impairment. JAMA. 2003;290(1):73-80. doi:10.1001/jama.290.1.41
Author Affiliations: Hebrew Rehabilitation Center for Aged Research and Training Institute, Department of Medicine, Beth Israel Deaconess Medical Center, and Division on Aging, Harvard Medical School, Boston, Mass (Dr Mitchell); and Center for Gerontology and Health Care Research, Department of Community Health, Brown Medical School, Providence, RI (Drs Teno, Roy, and Mor and Mr Kabumoto).
Context Empiric data and expert opinion suggest that use of feeding tubes is
not beneficial for older persons with advanced dementia. Previous research
has shown a 10-fold variation in this practice across the United States.
Objective To identify the facility and resident characteristics associated with
feeding tube use among US nursing homes residents with severe cognitive impairment.
Design, Setting, and Participants Cross-sectional study of all residents with advanced cognitive impairment
who had Minimum Data Set assessments within 60 days of April 1, 1999, (N =
186 835) and who resided in Medicare- or Medicaid-certified US nursing
Main Outcomes Measures Facility and resident characteristics described in the 1999 On-line
Survey Certification of Automated Records and the 1999 Minimum Data Set. Multivariate
analysis using generalized estimating equations determined the facility and
resident factors independently associated with feeding tube use.
Results Thirty-four percent of residents with advanced cognitive impairment
had feeding tubes (N = 63 101). Resident characteristics associated with
a greater likelihood of feeding tube use included younger age, nonwhite race,
male sex, divorced marital status, lack of advance directives, a recent decline
in functional status, and no diagnosis of Alzheimer disease. Controlling for
these patient factors, residents living in facilities that were for profit
(adjusted odds ratio [OR], 1.09; 95% confidence interval [CI], 1.06-1.12);
located in an urban area (OR, 1.14; 95% CI, 1.11-1.16); having more than 100
beds (OR, 1.04; 95% CI, 1.01-1.07); and lacking a special dementia care unit
(OR, 1.11; 95% CI, 1.07-1.15) had a higher likelihood of having a feeding
tube. Additionally, feeding tube use was more likely among residents living
in facilities that had a smaller proportion of residents with do-not-resuscitate
orders, had a higher prevalence of nonwhite residents, and lacked a nurse
practitioner or physician assistant on staff.
Conclusions More than one third of severely cognitively impaired residents in US
nursing homes have feeding tubes. Feeding tube use is independently associated
with both the residents' clinical characteristics and the nursing homes' fiscal,
organizational, and demographic features.
A growing proportion of the approximately 4 million older US adults with
Alzheimer disease or other dementias are now surviving to the advanced stages
of their illness. Eating and swallowing problems typically develop during
the terminal stages of dementias. Whether to initiate feeding tube use or
to focus on comfort is one of the most challenging dilemmas facing families,
clinicians, and institutions caring for these patients.1,2
The widespread use of feeding tubes among older persons with advanced
cognitive impairment in the United States3 is
concerning amid growing empirical data and expert opinion indicating that
feeding tube use has no demonstrable health benefits in this population and
may be associated with increased risks and discomfort.1,2 For
physician and/or institution practice to change, the factors that influence
feeding tube use among individuals with advanced cognitive impairment must
be understood. These influences may include the clinical characteristics of
the patient, the quality of the shared decision-making process, cultural attitudes
toward death and dying, and organizational features specific to the health
care environment in which the patient is receiving care.
The prevalence of feeding tube use varies considerably among nursing
home residents with advanced dementia who are living in different facilities,4 states,3,5 and
countries.6 These observations suggest that
broader influences, which are external to the patient, are important determinants
of this practice. A recent study found a 10-fold variation in feeding tube
use across the United States.3 However, only
state differences in the use of do-not-resuscitate (DNR) orders were found
to be associated with the practice, whereas regional laws governing hydration
and state Medicaid reimbursement policies were not. Therefore, additional
research is needed to understand the substantial variation in feeding tube
To date, investigations attempting to identify factors associated with
feeding tube use in this population have examined individual patient variables5,7- 9 or nursing
home characteristics, such as size, location, case-mix, staffing, and fiscal
organization, separately.4 However, these factors
do not operate independently. To address this limitation, we linked 2 large
national databases; 1 containing administrative, fiscal, and aggregated clinical
information on all licensed US nursing homes10 and
the second containing detailed demographic, functional, and clinical data
about the residents living in these facilities.11 Our
objective was to examine how patient factors and facility characteristics
independently influence feeding tube use among severely cognitively impaired
older persons residing in US nursing homes.
The study population was characterized using data from the 1999 National
Repository Resident Assessment Instrument Minimum Data Set (MDS).11 Data were derived from the MDS assessments completed
within 60 days of April 1, 1999, on nursing home residents living in all Medicare-
or Medicaid-certified US facilities during that period. Full MDS assessments
are completed within 2 weeks of nursing home admission, annually, and whenever
there is a significant change in status. Briefer MDS assessments are completed
on a quarterly basis. Only full MDS assessments were analyzed in this study.
We excluded residents whose reason for their MDS assessment was nursing home
admission. The MDS data for the remaining individuals represented random periods
in their nursing home stay.
Residents were included in the study sample if they had advanced cognitive
impairment as defined by a cognitive performance score (CPS) of six.12,13 The CPS uses 5 MDS variables to group
residents into 7 hierarchical cognitive performance categories. These categories
include 0 = intact; 1 = borderline intact; 2 = mild impairment; 3 = moderate
impairment; 4 = moderately severe impairment; 5 = severe impairment; and 6
= very severe impairment with eating problems. Residents identified as comatose
were excluded. The MDS was used to determine feeding tube use among residents.
The MDS does not distinguish between different types of feeding tubes.
Resident characteristics that were potentially associated with the use
of feeding tubes were selected from the data set a priori.5,7- 9 These
characteristics included age; sex; race or ethnicity; marital status; advance
directives (DNR order, living will, and durable power of attorney for health
care); diagnoses of Alzheimer disease, stroke, or cancer; and deterioration
in the ability to perform activities of daily living during the past 90 days
(based on a specific MDS variable). Age was grouped in 5-year ranges (<65,
65-69, 70-74, 75-79, 80-84, 85-89, and ≥90 years, which was the referent
category). Race or ethnicity was categorized as American Indian/Alaska Native,
Asian/Pacific Islander, black (not of Hispanic origin), Hispanic, and white
(not of Hispanic origin), which was the referent category. Marital status
was categorized as married, which was the referent category, never married,
separated, widowed, or divorced.
Facility characteristics were obtained from the MDS and the On-line
Survey Certification of Automated Records (OSCAR) data sets from 1999.10 OSCAR is a national database of information collected
annually as part of the nursing home survey and recertification process; it
contains data on facility demographics, fiscal and corporate structure, staffing,
and aggregated patient data. Facility characteristics that may be associated
with the use of feeding tubes among residents with advanced cognitive impairment
were selected a priori.4,9 All
licensed nursing homes in the United States were included in the analysis
(N = 15 135).
Facility variables were size (dichotomized at 100 beds, the median),
urban vs rural location, and fiscal characteristics and ownership including
chain membership, profit status, and whether Medicaid was the primary payer
for 80% or more of the beds in the facility. We also determined whether the
facility had a special care unit for dementia.
The level of staffing for nurses and certified nursing assistants was
defined as the number of full-time (35 h/wk) equivalents per bed. These variables
were categorized as less than 0.20 full-time equivalents, which was the referent
category, 0.21-0.40, 0.41-0.60, or higher than 0.60. Facilities were dichotomized
based on whether they had a full-time speech therapist employed on staff (at
least 35 h/wk) and a nurse practitioner or physician assistant on staff who
performed physician-delegated services.
The MDS data were used to create several aggregated patient variables
based on all residents in the facility—not just those in the study sample—to
describe specific nursing home characteristics that were unavailable in the
OSCAR data set. The proportion of residents in each facility with a DNR order
was grouped as less than 10%, 11%-20%, 21%-40%, 41%-60%, 61%-80%, and higher
than 80%, which was the referent category. The racial or ethnic makeup of
the facility was defined as the percentage of nonwhite residents and categorized
as 0%, which was the referent category, 0.1%-5.0%, 5.1%-10.0%, and higher
than 10%. The proportion of residents receiving intravenous therapy was included
to reflect the use of other invasive therapies by the facility and grouped
as 0%, 0.1%-1.0%, 1.1%-5.0%, 5.1%-10.0%, and higher than 10.0%, which was
the referent category. Finally, certain nursing homes may attract tube-fed
patients. To adjust for this factor, we determined the proportion of residents
with severe cognitive impairment admitted to each facility with feeding tubes
in 1999 and categorized this variable as less than 1.0%, which was the referent
category, 1.0%-5.0%, 5.1%-10.0%, and higher than 10.0%.
The dependent variable was whether a resident with advanced cognitive
impairment had a feeding tube. As such, the resident was the unit of analysis.
The independent variables were grouped as characteristics of individual residents
and characteristics of the facilities in which the residents with advanced
dementia lived. Descriptive statistics were used to present the proportion
of tube-fed and nontube-fed residents with each independent variable.
Multivariate analysis of the dichotomous outcome (the presence of a
feeding tube) was performed using logistic regression models with the generalized
estimating equation approach in SAS PROC GENMOD.14 The
generalized estimating equation method adjusts for the correlation among patients
residing in the same nursing facility. All the aforementioned resident and
facility characteristics were included as independent variables and entered
into the multivariate analysis. Therefore, the final logistic regression model
adjusts for all resident and facility characteristics. Adjusted odds ratios
(ORs) and 95% confidence intervals (CIs) were derived from these analyses.
When the outcome is not a rare event, as in this study, the OR may overestimate
the risk ratio. Therefore, the method by Zhang and Yu15 was
used to correct all ORs.
To describe the variability in feeding-tube practice in nursing homes
within and among states, the proportion of residents with advanced cognitive
impairment who had feeding tubes was determined and categorized based on terciles
of less than 16%, 16%-40%, and greater than 40%. The percentage of nursing
homes in each category was determined for all states. The states were also
ranked from those with greatest percentage of facilities in the highest tercile
(>40%) to those with the fewest nursing homes in that category. In addition,
the proportion of tube-fed residents in each state was determined.
A total of 186 835 nursing home residents in this nationwide sample
had a CPS score of 6 (very severe impairment with eating problems); of these,
33.8% (n = 63 101) had feeding tubes. The prevalence of feeding tube
use among residents in the other CPS categories is 6965 (2.0%) of 356 761
for a CPS score of 0 (intact); 6012 (3.1%) of 190 731 for a CPS score
of 1 (borderline intact); 6454 (3.2%) of 201 777 for a CPS score of 2
(mild impairment); 18 128 (4.5%) of 400 402 for a CPS score of 3
(moderate impairment); 15 141 (11.9%) of 127 275 for a CPS score
of 4 (moderately severe impairment); and 2619 (1.9%) of 135 964 for a
CPS score of 5 (severe impairment). Table
1 describes the facility and clinical characteristics of the residents. Table 2 presents the results of the multivariate
analysis, adjusting for all resident and facility characteristics.
Our multivariate analysis identified several resident characteristics
that were independently associated with feeding tube use in advanced cognitive
impairment. Residents who were younger and male were more likely to have feeding
tubes. The likelihood of feeding tube use was lowest among white residents.
Marital status was also independently associated with feeding tube use. Divorced
residents with advanced cognitive impairment had a greater likelihood of feeding
tube use compared with married residents.
A greater proportion of severely cognitively impaired residents with
feeding tubes did not have advance directives compared with residents without
feeding tubes. The lack of a DNR order, no durable power of attorney for health
care, and no living will were independently associated with feeding-tube status
in the multivariate model.
Residents without a diagnosis of Alzheimer disease were more likely
to have feeding tubes compared with other residents with advanced cognitive
impairment. A history of stroke was associated with a greater likelihood of
having a feeding tube, while residents with cancer were less likely to have
a feeding tube. Finally, a recent decline in functional status was independently
associated with feeding tube use.
Several demographic, organizational, and fiscal features of nursing
homes were independently associated with feeding tube use (Table 2). Residents in facilities lacking a dementia special care
unit had a greater likelihood of feeding tube use compared with residents
in facilities with these units. In addition, individuals were more likely
to have feeding tubes if they lived in nursing homes that were in located
in an urban setting, were run for profit, and that had more than 100 beds.
Chain ownership was not associated with feeding-tube status.
Residents in facilities with more than 80% of their beds dedicated to
Medicaid recipients were more likely to receive feeding tubes compared with
patients who resided in nursing homes with fewer beds dedicated to Medicaid
recipients. However, this relationship was not sustained after controlling
for other resident and facility characteristics. This association was confounded
primarily by the racial or ethnic profile of the facility, such that nursing
homes with a greater proportion of nonwhites were more likely to have more
than 80% of beds dedicated to Medicaid recipients and also to have residents
with feeding tubes.
Advance directives were independently associated with feeding-tube status
at the facility level. Older persons with advanced cognitive impairment residing
in nursing homes in which a relatively smaller proportion of the resident
population had DNR orders had a greater likelihood of having a feeding tube.
For example, feeding tube use was more likely in facilities in which the prevalence
of DNR orders was 61%-80% compared with facilities in which more than 80%
of residents had DNR orders (OR, 1.22; 95% CI, 1.13-1.30). This significant
trend persisted in a dose-response fashion as the frequency of DNR orders
The racial or ethnic profile of the facility was also independently
associated with the use of feeding tubes. Compared with nursing homes with
white residents only, feeding tube use was more likely to occur in facilities
with higher proportions of nonwhite residents.
The lack of a nurse practitioner or physician assistant on the nursing
home staff was independently associated with a higher likelihood of feeding
tube use among residents with advanced cognitive impairment (OR, 1.07; 95%
CI, 1.04-1.10). Other staffing variables, including the ratio of licensed
nurses and nursing assistants per bed and having a speech therapist on staff,
were not associated with the use of feeding tubes after multivariate adjustment.
Residents were more likely to receive feeding tubes if they lived in
facilities that had higher rates of new admissions with feeding tubes. The
facility rate of residents receiving intravenous therapy was not associated
with feeding-tube status.
Figure 1 presents the
facility rates of feeding tube use among nursing home residents with advanced
cognitive impairment in each state. Figure 1 highlights the variation in practice across facilities
both within and among states. For example, in Tennessee, the use of feeding
tubes among residents with advanced cognitive impairment is less than 16%
in 24% of facilities; ranges from 16%-40% in 42% of facilities; and exceeds
40% in 34% of facilities. The District of Columbia had the highest proportion
(90%) of nursing homes in which the use of feeding tubes exceeded 40%.
The number of nursing homes in each state and the proportion of tube-fed
residents with severe cognitive impairment in each state are presented in
the columns beside the Figure 1.
The prevalence of feeding tube use at the individual level was also the highest
in the District of Columbia where 64% of residents have feeding tubes.
This nationwide study demonstrates that the use of feeding tubes among
nursing home residents with severe cognitive impairment is common and associated
not only with their clinical features, but with the fiscal, organizational,
demographic, and ethnic or racial profile of the facility in which they live.
The relative influence of these nursing home characteristics is further demonstrated
by the observation that feeding tube practice varies considerably among facilities
within the same state. We identified several potentially modifiable factors
at the facility level that may reduce feeding tube use, including greater
use of advance directives, having a nurse practitioner or physician assistant
on staff, and having a special dementia care unit.
This study supports and extends earlier work examining factors associated
with feeding tube use in the institutionalized elderly with advanced cognitive
impairment.4,5,7- 9 However,
this investigation is unique in several important ways. First, the study includes
residents from all licensed US nursing homes. Second, it is the only study
to examine both resident and facility characteristics in a single analytic
model. Only one other investigation has comprehensively examined nursing homes'
characteristics associated with the use of feeding tubes in this population.4 However, this earlier study is limited because it
only examined facility characteristics associated with the overall rate of
feeding tube use at the facility level; it did not consider resident characteristics
in the model; data from only 6 states were included; and several important
facility factors were not analyzed. Finally, our use of the generalized estimating
equation approach to adjust for correlations among residents in the same nursing
home make the findings even more robust.
Using a national data set, our findings corroborate many previously
reported correlates of patient characteristics associated with feeding tube
use among individuals with advanced cognitive impairment including younger
age,5,8,9 no diagnosis
of Alzheimer disease,5,8,9 stroke,5,8,9 lack of advance directives
to limit aggressive care,5,8,9 and
nonwhite race.5,7,9 Some
studies, including ours, found an increase in feeding tube use among men,9 while others reported the opposite.5
Among the most notable observations in this study is the increased likelihood
of feeding tube use among residents living in for-profit nursing homes. This
observation is consistent with the notion put forward in the lay and scientific
press that feeding tube use among patients with advanced cognitive impairment
may be used by nursing homes as a means of cost-saving.16- 19 The
staff's time required for feeding residents by hand is expensive.18,19 In addition, Medicaid reimbursement
schemes in many states pay higher per diem rates for tube-fed residents compared
with similarly debilitated residents who do not have feeding tubes.18,20 However, Medicaid payments have been
shown not to be associated with feeding tube use at the state level.3 Moreover, the practice varies greatly among nursing
homes even within the same state. Therefore, while the potential for financial
incentives to favor use of feeding tubes exists, this association remains
to be proven.
Previous work has shown that the use of DNR orders is strongly associated
with feeding tube use in patients with advanced cognitive impairment at both
the patient5,8,9 and
state levels.3 This study underscores the important
role of DNR orders at the nursing home level.4 The
association between higher facility rates of DNR orders with a lower likelihood
of feeding tube use deserves further investigation. One potential explanation
is that facilities with a greater overall rate of DNR orders may be more proficient
at engaging surrogates in discussions that lead to decisions not to use a
feeding tube. In addition, families that request DNR orders may also be less
likely to want to have a feeding tube used. The MDS is unable to provide any
information about the decision-making process between families and health
This study demonstrates that the role of race or ethnicity in end-of-life
decision making extends beyond the background of the individual patient. It
has been widely reported that nonwhites tend to choose more aggressive end-of-life
care,21,22 including the use of
feeding tubes.5,7,9 Possible
explanations for this observation include different cultural attitudes toward
death and dying, apprehension of nonwhites toward the medical system, and
poor communication of advance directives to minorities by health care providers.
The racial or ethnic mix of the nursing home may be linked to other facility
factors that influence feeding tube use that we were unable to measure, such
as the background of the nursing home staff. Black physicians are less likely
to view feeding tube use in advanced dementia as a heroic measure.23
This study confirms that severely cognitively impaired residents living
in nursing homes that are larger,4 located
in urban areas,9 and lack dementia special
care units4 have a greater likelihood of using
feeding tubes. The closer proximity of urban nursing homes to tertiary medical
centers may translate into an increased use of high technology care at the
end of life.24 On the other hand, fewer feeding
tubes among nursing home residents with access to dementia special care units
suggests that these units may have greater success at guiding care toward
palliation. We also found that facilities that had either a nurse practitioner
or physician assistant on staff were less likely to have patients using feeding
tubes. Some data suggest that clinicians who practice solely in nursing homes
may be less likely to promote feeding tube use among residents with advanced
dementia compared with community-based clinicians.25
This study has several limitations that deserve comment. Some resident
and facility characteristics that may be associated with feeding tube use
were not available in the MDS and OSCAR data sets, such as religious affiliation,
characteristics of the nursing home staff, and facility-specific policies
regarding artificial nutrition and hydration. We also lacked information about
the quality of counseling between practitioners and families regarding the
use of feeding tubes in individuals with advanced cognitive impairment. In
addition, the MDS and OSCAR data may include some inaccuracies. However, any
inaccuracies would be likely to bias results toward the null. Finally, while
we were able to identify the pattern of feeding tube use in the United States
in 1999, it is possible that policies have since been introduced that have
changed the observed behavior.
The aggressiveness of care provided at the end of life is increasingly
recognized to be determined by more than the preferences and needs of individual
patients.24,26 This study confirms
that feeding tube use among older persons with advanced cognitive impairment
varies depending on the characteristics of the nursing home in which they
reside. As such, our findings highlight potential interventions and policy
changes at the facility level that could influence this practice. Comprehensive
implementation of advanced care planning is likely to reduce the use of feeding
tubes. However, future research will need to explore how the facilities' rates
of DNR orders relate to the broader process of shared decision making. Moreover,
if financial incentives are proven to favor feeding tube use, then reimbursement
policies that promote feeding by hand should be considered. Feeding tube use
may also be reduced by having providers and units dedicated to the care of
nursing home residents with advanced cognitive impairment. Finally, a greater
understanding of the influence of race or ethnicity on the use of feeding
tubes is needed to ensure informed and culturally sensitive decisions for