/clinical/,/clinical/cckm-tools/,/clinical/cckm-tools/content/,/clinical/cckm-tools/content/questionnaires/,/clinical/cckm-tools/content/questionnaires/related/,

/clinical/cckm-tools/content/questionnaires/related/name-97167-en.cckm

201606168

page

100

UWHC,UWMF,

Tools,

Clinical Hub,UW Health Clinical Tool Search,UW Health Clinical Tool Search,Questionnaires,Related

Estimating Prognosis for Nursing Home Residents With Advanced Dementia

Estimating Prognosis for Nursing Home Residents With Advanced Dementia - Clinical Hub, UW Health Clinical Tool Search, UW Health Clinical Tool Search, Questionnaires, Related


ORIGINAL CONTRIBUTION
Estimating Prognosis for Nursing Home
Residents With Advanced Dementia
Susan L. Mitchell, MD, MPH, FRCPC
Dan K. Kiely, MPH, MA
Mary Beth Hamel, MD, MPH
Pil S. Park, PhD
John N. Morris, PhD
Brant E. Fries, PhD
M
ANY FAMILIES AND HEALTH
care professionals in the
United States believe that
palliation is the most ap-
propriate goal for patients with end-
stage dementia.
1
However, many nurs-
ing home residents dying with advanced
dementia in the United States do not re-
ceive optimal palliative care.
2
Accurately estimating the life expec-
tancy of personswith advanceddemen-
tia is difficult andhinders palliative care
in this population.
3-6
Prognostic informa-
tion is important in guiding end-of-life
decisionmaking
7
and, in theUnitedStates,
for determininghospice eligibility.Medi-
carebeneficiariesmusthave anestimated
life expectancy of less than 6months to
be eligible for hospice. In a nationwide
survey, 80%ofhospices citeddifficulties
inpredicting survival as amajor problem
in the delivery of care to enrollees with
a primary diagnosis of dementia.
4,8
Less
than 1% of US hospice enrollees have a
primary diagnosis of dementia.
Fewstudieshave attempted todevelop
statisticalmodels to predict the 6-month
survival of personswith a primary diag-
nosis of advanced dementia.
3,5
The Na-
tional Hospice Organization eligibility
guidelines for patientswithdementia are
basedprimarilyon theFunctionalAssess-
ment Staging (FAST) criteria.
9,10
These
criteria havebeen criticizedbecause they
werenot derived fromempirical data, do
not accurately predict 6-month survival,
and cannot be applied to themajority of
patients with dementia whose disease
does not progress linearly.
3-6
Therefore, the objectives of this study
were to identify factors associated with
6-month mortality in newly admitted
nursing home residents with ad-
vanced dementia and to create a prac-
tical risk score to predict survival in this
population.
METHODS
Data Sources
The institutional review board at He-
brew Rehabilitation Center for Aged ap-
proved the conduct of this study.
Author Affiliations: Hebrew Rehabilitation Center
for Aged Research and Training Institute (Drs
Mitchell and Morris and Mr Kiely), the Department
of Medicine of Beth Israel Deaconess Medical Center
(Dr Mitchell), and Division on Aging, Harvard Medi-
cal School, Boston, Mass (Dr Mitchell); Division of
General Medicine and Primary Care, Beth Israel Dea-
coness Medical Center, Boston, Mass (Dr Hamel);
Institute of Gerontology (Drs Park and Fries) and
School of Public Health, University of Michigan (Dr
Fries), and Ann Arbor VA Medical Center, Ann
Arbor, Mich (Dr Fries).
Corresponding Author: Susan L. Mitchell, MD,MPH,
FRCPC, Hebrew Rehabilitation Center for Aged, 1200
Centre St, Boston, MA 02131 (smitchell
@mail.hrca.harvard.edu).
Context Survival varies for patients with advanced dementia, and accurate prog-
nostic tools have not been developed. A small proportion of patients admitted to hos-
pice have dementia, in part because of the difficulty in predicting survival.
Objectives To identify factors associated with 6-month mortality in newly admit-
ted nursing home residents with advanced dementia and to create a practical risk score
to predict 6-month mortality in this population.
Design, Setting, and Participants This was a retrospective cohort study of data
from theMinimumData Set (MDS). All Medicare or Medicaid licensed nursing homes
in New York and Michigan were included. Participants had advanced dementia and
were admitted to New York nursing homes between June 1, 1994, and December 30,
1998 (derivation cohort, n=6799), and to Michigan nursing homes from October 1,
1998, through July 30, 2000 (validation cohort, n=4631).
Main Outcome Measures MDS factors associated with 6-month mortality were
determined in the derivation group, and the resulting mortality risk score was evalu-
ated in the validation cohort. Risk score performance was compared with the cut point
of 7c on the Functional Assessment Staging (FAST) scale.
Results Among residents with advanced dementia, 28.3% (n=1922) died within 6
months of nursing home admission in the derivation cohort; 35.1% (n=1626) died in
the validation cohort. The 6-month mortality rate increased across risk scores (pos-
sible range, 0-19): 0 points, 8.9% mortality; 1 to 2, 10.8%; 3 to 5, 23.2%; 6 to 8,
40.4%; 9 to 11, 57.0%; and at least 12, 70.0% in the validation cohort. The area
under the receiver operating characteristic (AUROC) curve for predicting 6-month mor-
tality was 0.74 and 0.70 in the derivation and validation cohorts, respectively. Our
risk score demonstrated better discrimination to predict 6-month mortality (AUROC,
0.64 for a cutoff of �6 points vs 0.51 for FAST stage 7c).
Conclusion A risk score based on 12 variables from the MDS estimates 6-month
mortality for nursing home residents with advanced dementia with greater accuracy
than existing prognostic guidelines.
JAMA. 2004;291:2734-2740 www.jama.com
2734 JAMA, June 9, 2004—Vol 291, No. 22 (Reprinted) 2004 American Medical Association. All rights reserved.
Downloaded From: http://jama.jamanetwork.com/ by a University of Wisconsin -Madison User on 02/16/2015

The Omnibus Budget Reconcilia-
tion Act of 1987 requires that all Medi-
care- andMedicaid-certified nursing fa-
cilities in the United States periodically
conduct federally mandated standard-
ized, comprehensive assessments of all
residents by using the Resident Assess-
ment Instrument (RAI). The Mini-
mum Data Set (MDS) is the assess-
ment component of the RAI.
11-14
The MDS contains information on
each resident’s functional, medical, cog-
nitive, psychologic, and social status.
Assessments are required on admis-
sion, at quarterly intervals thereafter,
and whenever there is significant
change in resident status. Data are col-
lected by trained professionals (ie,
nurses, social workers, and therapists),
and each MDS item has its own ex-
plicit definition and coding conven-
tions. The interrater reliabilities for the
MDS items used in study were deter-
mined from data collected for clinical
purposes in comparable nursing homes
across the country.
12-14
The reliabili-
ties range from 0.50 to 0.99, with 6 vari-
ables having reliabilities below 0.70.
Themodel was derived by usingMDS
data from newly admitted residents to
all licensedMedicare orMedicaid nurs-
ing homes (n=634 facilities) in New
York between June 1, 1994, and De-
cember 30, 1998 (MDS version 1.0).
The model was validated
15
with MDS
data from newly admitted residents to
all licensedMedicare orMedicaid nurs-
ing homes (n=440 facilities) inMichi-
gan fromOctober 1, 1998, through July
30, 2000 (MDS version 2.0). The vari-
ables used in this study were defined
identically inMDS versions 1.0 and 2.0.
Population
The derivation and validation data sets
comprised persons aged at least 65 years
andwith advanced dementia whose rea-
son for their MDS assessment was nurs-
ing home admission. Advanced demen-
tia was defined as having a diagnosis of
dementia (Alzheimer disease or other
causes) and a Cognitive Performance
Score of 5 or 6.
16,17
The Cognitive Per-
formance Score uses 5 MDS variables
to group residents into 7 hierarchical
cognitive performance categories. A
Cognitive Performance Score of 5 or 6
(severe or very severe impairment with
eating problems) generally corre-
sponds to a Mini-Mental State Exami-
nation score of no more than 5.
16,17
Mortality Data
Mortality data and MDS information
werematched in the validation and deri-
vation data sets before the initiation of
this study. Slightly different methods
were used to identify residents who died.
In the derivation cohort, residents who
hadMDS information beyond 6months
of their admission date were consid-
ered to be alive. For the remaining resi-
dents, death information was obtained
from the National Death Index.
18
Matches were attempted on 11 vari-
ables to provide a complete match, par-
tial match, or no match with an indi-
vidual’s information. Residents in the
derivation cohort for whom a complete
match was found and whose death date
was within 6months of admission to the
nursing home were designated as hav-
ing died within 6months. Residents not
in the National Death Index or whose
death date was beyond 6months of their
admission date were considered alive at
6 months. Residents with partial or am-
biguousmatches were excluded from the
study. Of the 7014 residentsmeeting our
inclusion criteria, 215 residents (3.1%)
had partial matches in theNational Death
Index and were excluded from the
sample, leaving 6799 residents for in-
clusion in the analysis.
In the validation cohort, the Michi-
gan Death Registry was used to obtain
death data. All residents in the Michi-
gan MDS data set were linked to the
Michigan Death Registry by using the
8 variables. Satisfactory matches ob-
tained for residents in the validation co-
hort with a death date within 6months
of admission to the nursing homewere
designated as having died. If a satisfac-
tory match was not obtained, then the
resident was designated as alive.
Resident Characteristics
According to theNationalHospiceOrga-
nization guidelines,
9
related publica-
tions,
3-6,10,19-24
and clinical experience,
resident characteristics thatwere believed
to be associated with 6-month mortal-
ity were selected from the MDS assess-
ment completedwithin 21 days of nurs-
ing home admission. These variables
included demographic data, functional
status, diagnoses, and other health con-
ditions. Advance directives and specific
treatments (eg, tube feeding) were not
considered as independent variables
because our intentwas to develop a prog-
nostic model based on factors intrinsic
to the residents’ health status rather than
to examine how different management
strategies influence survival.
Demographic data included age, sex,
and race or ethnicity (white vs non-
white). Race and ethnicity are defined in
theMDS form. Other studies of survival
and dementia have included race
21
as a
prediction variable. Functional abilitywas
assessedwith the Activities of Daily Liv-
ing (ADL) scale (range=0, independent
in all, to 28, dependent in all).
25
A vari-
able that describes a resident as bedfast
most of the time was also included.
Diagnoses included in the analyses
were diabetes mellitus, congestive heart
failure, asthma or emphysema/chronic
obstructive pulmonary disease, cancer,
pneumonia or other respiratory tract in-
fection, cardiac dysrhythmias, any frac-
ture in the previous 180 days, urinary
tract infection in the previous 30 days,
and septicemia. Other health condi-
tions included were edema, hallucina-
tions or delusions, recurrent lung aspi-
rations, bowel incontinence, weight loss
(�5% in the previous 30 days or�10%
in the previous 180 days), dehydration,
insufficient fluids (did not consume al-
most all liquids in previous 3 days), fe-
ver, pressure ulcers (with at least some
loss of skin integrity), shortness of breath,
chewing or swallowing problems, no
more than 25% of food eaten at most
meals, not awakemost of the day (morn-
ing and afternoon), the need for oxy-
gen therapy in the previous 14 days, and
body mass index (BMI). We also exam-
ined whether residents were identified
as having an unstable condition, which
is defined in theMDS as a condition that
causes the resident’s usual cognitive,
PROGNOSIS FOR NURSING HOME RESIDENTS WITH DEMENTIA
2004 American Medical Association. All rights reserved. (Reprinted) JAMA, June 9, 2004—Vol 291, No. 22 2735
Downloaded From: http://jama.jamanetwork.com/ by a University of Wisconsin -Madison User on 02/16/2015

functional, or behavior pattern to fluc-
tuate, be precarious, or deteriorate.
Comparison With FAST
The FAST scale
10
is used to assess func-
tional change among patients with de-
mentia and consists of 7 major stages
with a total of 16 successive stages and
substages (possible range, stages 1-7f).
Stage 7 represents the most advanced
dementia and consists of 6 substages,
7a to 7f. These stages are defined as fol-
lows: 7a, speech is limited to 1 to 5
words; 7b, all intelligible vocabulary is
lost; 7c, nonambulatory; 7d, unable to
sit independently; 7e, unable to smile;
and 7f, unable to hold head up. Stage
7c of the FAST scale has been sug-
gested by the National Hospice Orga-
nization as an appropriate cutoff to en-
roll persons with a primary diagnosis
of dementia into hospice.
9
To be con-
sidered as stage 7c, patients must have
progressed through all the previous
stages of the FAST scale sequentially.
To compare the ability of our risk
score to predict 6-month survival with
that of FAST, variables were chosen from
theMDS thatmost closelymatch the de-
scription of FAST stages 6 through 7c
(TABLE 1). Residents with all of the fol-
lowing characteristics were considered
to be at FAST stage 7c: limited or more
extensive assistance needed for dress-
ing and toileting, supervision or more
assistance needed for bathing, urinary
and fecal incontinence at least twice a
week, rarely or never able tomake them-
selves understood, and inability to am-
bulate without extensive assistance.
Statistical Analyses
Derivation ofModel andRisk Score. Sur-
vival timewas the dependent variable for
all analyses. For residents who died
within 6months of nursing home admis-
sion, survival was defined as the dura-
tion between the admission and death
dates. Residents who did not die within
6 months of nursing home admission
were censored (considered alive). Inde-
pendent variables included all the afore-
mentioned resident characteristics.
Age, ADL score, and BMI were di-
chotomized to ease interpretation of the
Figure 1.Mortality Risk Index Score for Stratification of Residents Into Levels of Risk for
6-Month Mortality
Score Sheet to Estimate 6-Month Prognosis in Nursing Home Residents With Advanced Dementia
If Total Risk Score is…
0
1 or 2
3, 4, or 5
6, 7, or 8
9, 10, or 11
Risk Estimate of Death
Within 6 Months, %
8.9
10.8
23.2
40.4
57.0
≥12 70.0
Risk Factor From Minimum Data Set Points Score
Activities of Daily Living Scale = 28

1.9 –––––
Male Sex 1.9 –––––
Cancer 1.7 –––––
Oxygen Therapy Needed in Prior 14 Days 1.6 –––––
Congestive Heart Failure 1.6 –––––
Shortness of Breath 1.5 –––––
<25% of Food Eaten at Most Meals 1.5 –––––
Unstable Medical Condition 1.5 –––––
Bowel Incontinence 1.5 –––––
Bedfast 1.5 –––––
Age >83 y 1.4 –––––
Not Awake Most of the Day 1.4 –––––
Total Risk Score, Rounded to Nearest Integer
Possible Range, 0-19

The Activities of Daily Living Scale is obtained by summing the resident's self-
performance ratings on the Minimum Data Set for the following 7 functional activities:
bed mobility, dressing, toileting, transfer, eating, grooming, and locomotion. In the
Minimum Data Set, functional ability is rated on 5-point scale for each activity
(0, independent; 1, supervision; 2, limited assistance; 3, extensive assistance; and
4, total dependence). A total score of 28 represents complete functional dependence.
Table 1. Description of Functional Assessment Stages and Comparable Minimum Data Set
Variables
Functional Assessment Stage Minimum Data Set Variable
6a = Improperly putting on clothes without
assistance/cueing occasionally or more
frequently over the past weeks
Limited or more extensive assistance
required to dress on at least several
occasions during the last 7 days
6b = Unable to bathe properly (eg, difficulty
adjusting water temperature)
occasionally or more frequently over the
past weeks
Supervision or more assistance required to
bathe during the last 7 days
6c = Inability to handle the mechanics of
using the toilet occasionally or more
frequently over the past weeks
Limited or more extensive assistance
required to use the toilet on at least
several occasions during the last 7 days
6d = Urinary incontinence occasionally or
more frequently over the past weeks
Urinary incontinence at least twice a week
6e = Bowel incontinence occasionally or
more frequently over the past weeks
Bowel incontinence at least twice a week
7a = Ability to speak limited to �1 intelligible
word in an average day
Rarely/never makes self understood
7b = All intelligible vocabulary is lost Rarely/never makes self understood
7c = Nonambulatory Extensive assistance (or total dependence)
required for locomotion (ie, move
between locations) during the last 7 days
PROGNOSIS FOR NURSING HOME RESIDENTS WITH DEMENTIA
2736 JAMA, June 9, 2004—Vol 291, No. 22 (Reprinted) 2004 American Medical Association. All rights reserved.
Downloaded From: http://jama.jamanetwork.com/ by a University of Wisconsin -Madison User on 02/16/2015

hazard ratio and to simplify themortal-
ity risk score. Age was dichotomized at
the median (�83 years); BMI was di-
chotomized at no more than 21 be-
cause this cutoff is below the range con-
sidered normal by standard clinical
nutritional assessment
26
and because it
also represented themedian value. ADL
score was dichotomized at 28 because
this value represented complete func-
tional dependence andwas the 75th per-
centile in the distribution of ADL scores
for the derivation cohort. Analysis of
these characteristics as dichotomous vs
continuous variables did not change the
prognostic power of the final survival
model’s comparable c statistic.
Unadjusted associations were ana-
lyzed with Cox proportional hazards
models (SAS Institute Inc, Cary, NC; ver-
sion 6.11). Characteristics that were sig-
nificantly (P�.05) associated with sur-
vival in these unadjusted analyses were
entered into Cox proportional hazards
regression model in a stepwise fashion.
To derive themost parsimoniousmodel
and because of the large sample size, only
variables associatedwith survival at a sig-
nificance level P�.001 after the step-
wise procedure were retained in the
model. Including variables not meeting
this level of significance did not im-
prove the predictive ability of the final
model. Hazard ratios, 95% confidence in-
tervals, and P values were derived from
these analyses. Proportional hazards as-
sumptions were met (time interaction
terms for the covariates had a P�.05).
Amortality risk index score was cre-
ated to stratify residents into different
levels of risk for 6-month mortality
(FIGURE 1). A point value was as-
signed to each characteristic accord-
ing to the hazard ratios in the final mul-
tivariate model from the derivation
cohort. Point values were summed for
all mortality-related characteristics pres-
ent for each resident and rounded to the
nearest integer (values of 0.5 were
rounded up to the next highest
integer).
15
Risk scores ranged from 0 to
19, and 6 risk categories were created
by combining risk scores with similar
mortality rates in the derivation co-
hort (0, 1-2, 3-5, 6-8, 9-11, and�12).
Validation of the Model. We ap-
plied the mortality risk score devel-
oped by using the derivation cohort to
the validation cohort and determined
the proportion of residents in the vali-
dation cohort who died in each risk cat-
egory, as defined by the hazard ratios
obtained in the derivation cohort.
Operating Characteristics of the
Model and Risk Score. To examine the
discrimination of our risk score, we de-
termined the c statistic, representing the
area under the receiver operating char-
acteristic (AUROC) curve.
27
We calcu-
lated the c statistic for the derivation and
validation cohorts by using logistic re-
gression, with 6-monthmortality as the
outcome and the risk score as the sole
independent variable. Testing in the vali-
dation cohort was based on the risk score
created from the derivation cohort.
Wealsoused logistic regression tode-
termine the c statistic for FAST stage 7c
simulatedwithMDSvariables to estimate
the ability of FAST criteria to predict
6-monthmortality in the validation co-
Table 2. Characteristics of Residents With Advanced Dementia and Their Associations With
6-Month Mortality in the Derivation Cohort (n = 6799)
Characteristic No. (%) of Residents Unadjusted HR (95% CI)
Demographic
Age �83 years, median 3075 (45.2) 1.5 (1.4-1.7)
Male sex 2257 (33.2) 1.8 (1.6-1.9)
Nonwhite race/ethnicity 1366 (20.1) 1.2 (1.1-1.4)
Functional status
Activities of daily living score = 28* 1747 (25.7) 2.5 (2.3-2.8)
Bedfast 523 (7.7) 2.6 (2.3-3.0)
Diagnosis
Diabetes mellitus 1113 (16.4) 1.4 (1.2-1.5)
Congestive heart failure 958 (14.1) 2.1 (1.9-2.3)
Asthma or emphysema/COPD 520 (7.6) 1.6 (1.4-1.8)
Cancer 575 (8.5) 2.1 (1.8-2.4)
Pneumonia or respiratory tract
infection
641 (9.4) 1.8 (1.6-2.0)
Cardiac dysrhythmia 768 (11.3) 1.4 (1.2-1.6)
Any fracture in the previous 180 days 594 (8.7) 1.0 (0.9-1.2)
Urinary tract infection 1343 (19.8) 1.3 (1.1-1.4)
Septicemia 53 (0.8) 1.8 (1.2-2.7)
Other health conditions
Edema 1155 (17.0) 1.5 (1.3-1.7)
Hallucinations or delusions 181 (2.7) 1.0 (0.7-1.3)
Aspiration 107 (1.6) 2.1 (1.6-2.8)
Bowel incontinence 5334 (78.4) 2.3 (2.0-2.7)
Recent weight loss 1074 (15.8) 1.9 (1.7-2.1)
Dehydration 370 (5.4) 1.7 (1.4-2.0)
Insufficient fluid intake 742 (10.9) 1.6 (1.4-1.8)
Fever 533 (7.8) 2.1 (1.8-2.4)
Pressure ulcers 1159 (17.0) 1.7 (1.5-1.9)
Shortness of breath 240 (3.5) 3.6 (3.0-4.3)
Unstable medical conditions 1626 (23.9) 1.9 (1.8-2.1)
Chewing or swallowing problem 2331 (34.3) 1.8 (1.6-1.9)
�25% of food eaten at most meals 2458 (36.2) 1.5 (1.4-1.6)
Not awake most of day 474 (7.0) 2.1 (1.8-2.4)
Body mass index �21, median† 3404 (50.1) 1.4 (1.3-1.5)
Oxygen therapy in prior 14 days 565 (8.3) 3.1 (2.7-3.5)
Abbreviations: CI, confidence interval; COPD, chronic obstructive pulmonary disease; HR, hazard ratio.
*Total activities of daily living score (0-28) is the sum of scores in each of 7 domains of function, including bed mobility,
dressing, using the toilet, transfer, eating, grooming, and locomotion. Each is scored on a 5-point scale (0, inde-
pendent; 1, supervision; 2, limited assistance; 3, extensive assistance; and 4, total dependence). A score of 28 rep-
resents complete functional dependence.
†Body mass index was calculated as weight in kilograms divided by the square of height in meters.
PROGNOSIS FOR NURSING HOME RESIDENTS WITH DEMENTIA
2004 American Medical Association. All rights reserved. (Reprinted) JAMA, June 9, 2004—Vol 291, No. 22 2737
Downloaded From: http://jama.jamanetwork.com/ by a University of Wisconsin -Madison User on 02/16/2015

hort. A c statistic for a single cut point on
our scorewas calculated in the validation
set to compare with FAST stage 7c.
To examine the practical application
of using the risk score to predict 6-month
mortality, we calculated the sensitivity,
specificity, andpositive andnegative pre-
dictive values for various cut points in
the risk score to predict 6-month mor-
tality for the derivation and validation
cohorts. We generated ROC curves for
the derivation and validation cohorts
according to these analyses.
RESULTS
Risk Score Derivation
The derivation cohort consisted of 6799
residents with advanced dementia who
were admitted to New York nursing
homes during the study period, of
whom 28.3% (n=1922) died within
6 months.
TABLE 2 describes the characteris-
tics of the residents in the derivation co-
hort and the unadjusted associations be-
tween each independent variable and
survival. Only fracture and hallucina-
tions or delusions were not signifi-
cantly associated with survival. All other
variables were entered into the step-
wisemultivariate Cox proportional haz-
ards model. The final model included
the following variables: ADL score of
28, male sex, cancer, the need for oxy-
gen therapy, congestive heart failure,
shortness of breath, nomore than 25%
of food eaten at most meals, an un-
stable condition, bowel incontinence,
bedfast, older than 83 years, and not
awake most of the day (TABLE 3).
Using the hazard ratios in the final
multivariate model, we calculated a risk
score for each resident. For example,
an 85-year-old male resident with ad-
vanced dementia who was bedfast and
totally functionally dependent (ADL
score, 28), had congestive heart fail-
ure, and used oxygen had a risk score
of 1.4+1.9 +1.5+1.9+1.6+1.6=9.9, or
10 after rounding to the nearest inte-
ger. Risk scores with similar mortality
rates were combined to produce 6 risk
categories. TABLE 4 shows by category
the proportion and number of resi-
dents in the derivation cohort who died
and also presents the percentage of resi-
dents in the total sample within each
risk category. Those with risk scores of
3, 4, and 5 represented 46.2% of the
derivation data set.
Risk Score Validation
In the validation cohort, 4631 older
persons with advanced dementia
were admitted to Michigan nursing
homes, of whom 35.1% (n=1626)
died within 6 months. Table 4 pre-
sents the number and proportion of
residents in each risk category in the
validation cohort and the proportion
of those who died. Risk categories
were calculated by using the hazard
ratios from the derivation set. The
mortality rates in each category were
similar to those in the derivation
cohort, except for a higher propor-
tion of residents dying with a score
of 0 and a lower proportion dying
with risk scores of at least 12 (Table
4). Those with risk scores of 3, 4,
and 5 represented 37.9% of the vali-
dation cohort.
Operating Characteristics
The c statistic representing the AUROC
curve was 0.74 in the derivation co-
hort and 0.70 in the validation cohort
(FIGURE 2).
TABLE 5 presents the operating char-
acteristics of the risk score at various
cut points to predict 6-month mortal-
ity in the validation and derivation co-
horts. A potential application of this ap-
proach would be to consider severely
demented residents with risk scores
above specific cut points as eligible for
hospice services while excluding resi-
dents with lower scores. For example,
if hospice eligibility included only resi-
dents with risk scores of 9 or higher,
then according to the validation data,
59.7% of enrollees would die within 6
months of admission (positive predic-
tive value), but only 28.7% of resi-
dents with advanced dementia who died
within that period would be eligible for
the program (sensitivity).With broader
inclusion criteria, for example, a risk
score cutoff of 6 or higher, only 46.6%
of enrollees would die within 6months,
and 72.8% of residents with advanced
dementia who died within that time
would be eligible for hospice care.
Table 3. Multivariate Proportional Hazards
Model of Characteristics Associated With
6-Month Mortality Among Residents
With Advanced Dementia in the Derivation
Cohort (n = 6799)
Characteristic
HR
(95% CI)
Activities of daily living
score = 28*
1.9 (1.7-2.1)
Male sex 1.9 (1.7-2.1)
Cancer 1.7 (1.5-1.9)
Oxygen therapy 1.6 (1.4-1.8)
Congestive heart failure 1.6 (1.4-1.7)
Shortness of breath 1.5 (1.3-1.9)
�25% Food eaten 1.5 (1.4-1.7)
Unstable medical conditions 1.5 (1.3-1.6)
Bowel incontinence 1.5 (1.3-1.7)
Bedfast 1.5 (1.3-1.7)
Age �83 years, median 1.4 (1.3-1.6)
Not awake most of day 1.4 (1.2-1.6)
Abbreviations: CI, confidence interval; HR, hazard ratio.
*Total activities of daily living score (0-28) is the sum of scores
in each of 7 domains of function including bed mobility,
dressing, using the toilet, transfer, eating, grooming, and
locomotion. Each is scored on a 5-point scale (0, inde-
pendent; 1, supervision; 2, limited assistance; 3, exten-
sive assistance; and 4, total dependence). A score of 28
represents complete functional dependence.
Table 4. Proportion of Residents With Advanced Dementia in Each Risk Category and Those
Who Died Within 6 Months of Nursing Home Admission in Derivation (n = 6799) and
Validation Cohorts (n = 4631)
Risk Score*
Total No. (%) Residents No. (%) Residents Who Died
Derivation Cohort Validation Cohort Derivation Cohort Validation Cohort
0† 272 (4.0) 56 (1.2) 4 (1.5) 5 (8.9)
1-2 938 (13.8) 278 (6.0) 79 (8.4) 30 (10.8)
3-5 3141 (46.2) 1755 (37.9) 666 (21.2) 407 (23.2)
6-8 1795 (26.4) 1727 (37.3) 732 (40.8) 698 (40.4)
9-11 517 (7.6) 648 (14.0) 332 (64.2) 369 (57.0)
�12 136 (2.0) 167 (3.6) 109 (80.1) 117 (70.0)
*Risk score is based on point assignments from hazard ratios obtained from the derivation cohort.
†Residents with a risk score of zero did not have any of the characteristics included in the final prognostic model.
PROGNOSIS FOR NURSING HOME RESIDENTS WITH DEMENTIA
2738 JAMA, June 9, 2004—Vol 291, No. 22 (Reprinted) 2004 American Medical Association. All rights reserved.
Downloaded From: http://jama.jamanetwork.com/ by a University of Wisconsin -Madison User on 02/16/2015

Application of FAST
In the validation cohort, 20.2% (n=937)
of residentswith advanced dementiamet
the simulatedMDScriteria for FAST stage
7c, ofwhom38.5%diedwithin 6months
of nursing home admission (positive pre-
dictive value). Only 22.2% of residents
who died within 6 months met the cri-
teria for FAST stage 7c (sensitivity).
Whenweused logistic regression to pre-
dict death at 6months, the c statistic for
the simulated FAST stage of 7cwas 0.51
in the validation cohort, indicating poor
discrimination. In comparison, the c sta-
tistic using the single cut point of at least
6 in our risk score was 0.64.
COMMENT
In this study, we have derived and vali-
dated amodel to predict 6-monthmor-
tality for nursing home residents with
advanced dementia. Our risk score of-
fers an improvement over existing prog-
nostic guidelines used in this popula-
tion because it is based on empiric data,
has greater predictive power, and uses
standardized, readily available MDS
assessments.
The few investigations that have spe-
cifically identified factors associatedwith
survival in advanced dementia differ
from our study in important ways, so
comparisons are limited.
3,5,19-21,23
None-
theless, our study confirms that the fol-
lowing characteristics are associatedwith
poorer survival in advanced dementia:
older age,
3,21,23
greater functional im-
pairment,
3,5,19,21,23
male sex,
21,23
cardio-
vascular disease,
21
diabetes mellitus,
21
and poor nutritional status.
5,21
These fac-
tors are also associated with increased
mortality in dementia, regardless of the
stage.
21,23,29
We also found that the risk
factors for death among nursing home
residents with advanced dementia were
similar to those of the general nursing
home population,
22
of which a substan-
tial proportion have dementia. The need
for oxygen therapy and not being awake
most of the day were the only addi-
tional factors that we identified to be spe-
cifically associated with survival in ad-
vanced dementia.
Earlier work demonstrates the chal-
lenge of estimating short-term progno-
sis among patients with advanced de-
mentia.
3,5,20
Our risk score demonstrates
moderately good power to predict
6-month survival among newly admit-
ted nursing home residents with ad-
vanced dementia. It performed better
than stage 7c of FAST when simulated
with MDS variables, which had a pre-
dictive ability that was equal to chance.
Moreover, in a study involving 47 hos-
pice enrollees with dementia, 41% of
enrollees could not be staged by using
FAST criteria because their disease
had not progressed in the ordinal se-
quence of the scale.
5
Advanced dementia is an incurable,
progressive condition for which pallia-
tion is often the primary goal of care,
regardless of life expectancy. Al-
though our model predicted 6-month
survival in advanced dementia with
greater accuracy than available prog-
nostic systems, these analyses high-
light the practical limitations of using
prognostic estimates as criteria to de-
termine access to palliative care ser-
vices. For example, with respect to the
Medicare hospice program, narrow eli-
gibility criteria ensure that the major-
ity of enrollees will die within 6months
but exclude a substantial proportion of
persons with advanced dementia who
also die during that period. With
broader inclusion criteria, a greater pro-
portion of patients who die within 6
months would be eligible for hospice
services, but a larger percentage of en-
rollees would survive beyond 6months.
Similar problems were demonstrated
when prognostic criteria from the Study
to Understand Prognoses and Prefer-
ences for Outcomes and Risks of Treat-
ment study were used to determine
hospice eligibility for seriously ill hos-
pitalized patients with other noncan-
cer diagnoses.
28
More restrictive inclu-
sion criteria would be an acceptable
approach provided that high-quality
palliative care was available to all resi-
dents with advanced dementia within
the existing framework of comprehen-
sive nursing homemanagement. Alter-
natively, broader eligibility criteria
would be a reasonable strategy if
hospice were willing to enroll persons
Figure 2. Receiver Operating Characteristic
(ROC) Curves for Risk Score’s Prediction of
6-Month Mortality in the Derivation
(n=6799) and Validation Cohorts (n=4631)
100
60
40
80
20
0
0 20 60 80 10040
100 80 60 2040 0
1– Specificity, %
Specificity, %
S
e
n
s
i
t
i
v
i
t
y
,

%
Derivation Cohort
Validation Cohort
The area under the ROC curve is 0.74 in the deriva-
tion cohort and 0.70 in the validation cohort.
Table 5. Operating Characteristics of Selected
Risk Score Cutoffs to Predict 6-Month
Mortality Following Nursing Home Admission
for Residents With Advanced Dementia
Risk Score
Cutoff
Derivation
Cohort
(n = 6799)
Validation
Cohort
(n = 4631)
Sensitivity*
�1 99.8 99.7
�3 95.7 97.8
�6 61.0 72.8
�9 22.9 28.7
�12 5.8 7.2
Specificity†
�1 5.5 1.7
�3 23.1 9.9
�6 73.8 54.8
�9 95.6 88.8
�12 99.4 98.3
Positive Predictive Value‡
�1 29.4 35.4
�3 32.9 37.0
�6 47.9 46.6
�9 67.4 59.7
�12 79.7 70.2
Negative Predictive Value§
�1 98.5 91.2
�3 93.1 89.5
�6 82.8 78.8
�9 75.9 69.9
�12 72.8 66.2
*Sensitivity: proportion of residents who died within 6
months of admission with a risk score above cut point.
†Specificity: proportion of residents who survived beyond
6months of admission with a risk score below cut point.
‡Positive predictive value: proportion of residents with a
risk score above the cut point who died within 6 months
of admission.
§Negative predictive value: proportion of residents with a
risk score below the cut point who survived beyond 6
months of admission.
PROGNOSIS FOR NURSING HOME RESIDENTS WITH DEMENTIA
2004 American Medical Association. All rights reserved. (Reprinted) JAMA, June 9, 2004—Vol 291, No. 22 2739
Downloaded From: http://jama.jamanetwork.com/ by a University of Wisconsin -Madison User on 02/16/2015

needing palliative care for longer than
6 months.
This study has some limitations that
deserve comment. First, the prognostic
model was derived in a population of
older persons with advanced dementia
who were recently admitted to a nurs-
ing home. Therefore, our risk scoremay
not be generalizable to residents who
have lived in nursing homes for longer
periods or for those living in the com-
munity. Second, by using admission data
to predict death within 6 months, we
could not account for changes in health
status during the interim period thatmay
influence survival. Althoughwe consid-
ered all acute illnesses available in the
MDS data set as independent variables,
it is possible that other factors associ-
ated with high short-term mortality in
advanced dementia were unavailable for
analysis. Moreover, there may not have
been adequate power to demonstrate sta-
tistically significant associations be-
tween uncommon conditions (eg, sep-
ticemia) and 6-month mortality. Third,
the risk score was derived and vali-
dated with data collected retrospec-
tively. Prospective validation would be
helpful to further assess the usefulness
of the risk score in clinical practice.
29
Fi-
nally, despite our best efforts to define
FAST stage 7c by using MDS variables,
our simulation closely approximates but
does not replicate the original scale.
Recent work indicates that the me-
dian survival after the onset of symp-
toms of dementia is shorter than previ-
ously estimated (3-6 years),
30,31
underscoring the need to plan for the end
stage of this illness. High-quality pallia-
tive care should be available to the large
proportion of persons with advanced de-
mentia who will be cared for in nursing
homes. Determining the best way to pro-
vide that care deserves the attention of
health care providers and policy mak-
ers. If hospice eligibility continues to re-
quire a high likelihood of death within
6 months, then the majority of patients
with advanced dementia in nursing
homes will not receive hospice ser-
vices. Therefore, alternative strategies to
deliver comprehensive palliative care to
this population should be sought.
32,33
While these issues are debated, the risk
score derived in this study offers a prac-
tical approach for estimating with rea-
sonable accuracy the 6-month progno-
sis of older nursing home residents with
advanced dementia.
Author Contributions: DrMitchell had full access to all
of the data in the study and takes responsibility for the
integrity of the data and accuracy of the data analysis.
Study concept and design:Mitchell, Hamel.
Acquisition of data:Morris, Fries.
Analysis and interpretation of data: Mitchell, Kiely,
Hamel, Park, Morris, Fries.
Drafting of the manuscript: Mitchell, Kiely, Hamel,
Morris, Fries.
Critical revision of the manuscript for important in-
tellectual content:Mitchell, Kiely, Hamel, Park, Morris,
Fries.
Statistical expertise:Mitchell, Kiely, Hamel, Park, Fries.
Obtained funding:Mitchell, Morris.
Administrative, technical, or material support:Mitchell,
Morris, Fries.
Supervision: Hamel, Morris, Fries.
Funding/Support: This work was supported by the He-
brew Rehabilitation Center for Aged (HRCA) Re-
search and Training Institute, the Marcus Apple-
baum Fund at the HRCA, a Teaching Nursing Home
Award (AG04390) and the Harvard Older American
Independence Center Grant (AG08812) from the Na-
tional Institute on Aging, Bethesda, Md (Drs Mitchell
and Morris). Dr Mitchell is supported by the NIH-
NIAMentored Patient-Oriented Research Career De-
velopment Award (K23AG20054). Dr Hamel is a re-
cipient of a Paul Beeson Physician Faculty Scholar Aging
Research Award.
Role of the Sponsors: The funding sources for this
study played no role in the design or conduct of the
study; the collection, analysis, interpretation, or prepa-
ration of the data; or in the preparation, review, or
approval of the manuscript.
REFERENCES
1. Luchins DJ, Hanrahan P.What is appropriate health
care for end-stage dementia? J AmGeriatr Soc. 1993;
41:25-30.
2. Mitchell SL, Kiely DK, Hamel MB. Dying with ad-
vanced dementia in the nursing home. Arch Intern
Med. 2004;164:321-326.
3. Volicer BJ, Hurley A, Fabiszewski KJ, Montgom-
ery P, Volicer V. Predicting short-term survival for pa-
tients with advanced Alzheimer’s disease. JAmGeri-
atr Soc. 1993;41:535-540.
4. Hanrahan P, Luchins DJ. Access to hospice pro-
grams in end-stage dementia: a national survey of hos-
pice programs. J Am Geriatr Soc. 1995;43:56-59.
5. Luchins DJ, Hanrahan P,Murphy K. Criteria for en-
rolling dementia patients in hospice. J AmGeriatr Soc.
1997;45:1054-1059.
6. Volicer L. Hospice care for dementia patients. JAm
Geriatr Soc. 1997;45:1147-1149.
7. Weeks JC, Cook EF, O’Day SJ, et al. Relationship be-
tween cancer patients’ predictions of prognosis and their
treatment preferences. JAMA. 1998;279:1709-1714.
8. Christakis NA, Escarce JJ. Survival of Medicare pa-
tients after enrollment in hospice programs. N Engl J
Med. 1996;335:172-178.
9. National Hospice Organization. Medical Guide-
lines for Determining Prognosis in Selected Non-
Cancer Diseases. 2nd ed. Arlington, Va: National Hos-
pice Organization; 1996.
10. Reisberg B. Functional assessment staging (FAST).
Psychopharmacol Bull. 1988;24:653-659.
11. Morris JN, Hawes C, Fries BE, et al. Designing the
National Assessment Instrument for Nursing Homes.
Gerontologist. 1990;30:293-307.
12. Hawes C, Morris JN, Phillips CD, Mor V, Fries BE,
Nonemaker S. Reliability estimates for theMinimumData
Set for nursing home resident assessment and care
screening (MDS). Gerontologist. 1995;35:172-178.
13. Morris JN, Nonemaker S,Murphy K, Hawes C, Fries
BE, Phillips C. A commitment to chance: revision of HF-
CA’s RAI. J Am Geriatr Soc. 1997;45:1011-1016.
14. Mor V, Angelelli J, Jones R, Roy J, Moore T, Mor-
ris JN. Inter-rater reliability of nursing home quality
indicators in the US [BMC Health Services Research
Web site]. Available at: http://www.biomedcentral
.com/1472-6963/3/20. Accessed February 23, 2004.
15. Wasson JH, Sox HC, Neff RK, Goldman L. Clini-
cal prediction rules: applications and methodological
standards. N Engl J Med. 1985;313:793-799.
16. Morris JN, Fries BE,Mehr DR, et al.MDSCognitive
Performance Scale. J Gerontol. 1994;49:M174-182.
17. Hartmaier SL, Sloane PD, Guess HA, Koch GG,
Mitchell CM, Phillips CD. Validation of the Minimum
Data Set Cognitive Performance Scale: agreement with
Mini-Mental State Examination. J Gerontol A Biol Sci
Med Sci. 1995;50:M128-M133.
18. Bilgrad R. National Death Index User’s Manual.
Hyattsville, Md: Dept of Health and Human Services,
Centers for Disease Control and Prevention, Na-
tional Center for Health Statistics; 2000.
19. CarlsonMC, Brandt J, Steele C, Baker A, Stern Y,
Lyketsos CG. Predictor index of mortality in demen-
tia patients upon entry into long-term care. J Geron-
tol A Biol Sci Med Sci. 2001;56:M567-M570.
20. Fabiszewski KJ, Volicer B, Volicer L. Effect or an-
tibiotic treatment on the outcome of fevers in insti-
tutionalized Alzheimer patients. JAMA. 1990;263:
3168-3172.
21. Gambassi G, Landi F, Lapane KL, Sgadari A, Mor
V, Bernabei R. Predictors of mortality in patients with
Alzheimer’s disease living in nursing homes. J Neurol
Neurosurg Psychiatry. 1999;67:59-65.
22. Flacker JM, Kiely DK.Mortality-related factors and
1-year survival in nursing home residents. JAmGeri-
atr Soc. 2003;51:213-221.
23. Aquero-Torres H, Fratiglioni L, Guo Z, Viitanen
M, Winblad B. Prognostic factors in very old de-
mented adults: a seven-year follow-up from a popu-
lation-based survey in Stockholm. J Am Geriatr Soc.
1998;46:444-452.
24. van Dijk PTM, Dippel DWJ, Habbema JD. Sur-
vival of patients with dementia. J AmGeriatr Soc. 1991;
39:603-610.
25. Morris JN, Fries BE, Morris SA. Scaling ADLs with
the MDS. J Gerontol A Biol Sci Med Sci. 1999;54:
M546-M553.
26. Committee on Diet and Health, Food and Nutri-
tion Board, Commission on Life Sciences, National Re-
search Council on Diet and Health. Implications for
Reducing Chronic Disease Risk. Washington, DC: Na-
tional Academy Press; 1989.
27. Hanley JA, McNeil BJ. The meaning and use of
the area under a receiver operating characteristic (ROC)
curve. Radiology. 1982;143:29-36.
28. Fox E, Landrum-McNiff K, Zhong Z, Dawson NV,
WuAW, Lynn J. Evaluating prognostic criteria for deter-
mininghospice eligibility in patientswith advanced lung,
heart or liver disease. JAMA. 1999;282:1638-1645.
29. Laupacis A, Sekar N, Stiell IG. Clinical prediction
rule: a review and suggested modifications of meth-
odological standards. JAMA. 1997;277:488-494.
30. Wolfson C, Wolfson DB, Asgharian M, et al. A re-
evaluation of the duration of survival after the onset of
dementia. N Engl J Med. 2001;344:1111-1116.
31. Larson EB, ShadlenM,Wang L, et al. Survival af-
ter the diagnosis of Alzheimer’s disease. Ann Intern
Med. 2004;140:501-509.
32. Lynn J. Caring at the end of our lives. N Engl J
Med. 1996;335:172-178.
33. Berger A. Palliative care in long-term care facili-
ties: a comprehensive model. J AmGeriatr Soc. 2001;
49:1570-1571.
PROGNOSIS FOR NURSING HOME RESIDENTS WITH DEMENTIA
2740 JAMA, June 9, 2004—Vol 291, No. 22 (Reprinted) 2004 American Medical Association. All rights reserved.
Downloaded From: http://jama.jamanetwork.com/ by a University of Wisconsin -Madison User on 02/16/2015