LOW-CARE NURSING HOME RESIDENTS
M293
The goal of our study was to develop an improved clinical
understanding of low-ADL–dependent long-staying nursing
home residents (i.e., few ADL impairments and more than 2
months as a nursing home resident) and their clinical course
over time. We looked for clinical (medical and psychiatric)
as well as social and behavioral characteristics that might ac-
count for their presence in the nursing home despite rela-
tively preserved ADL function. In addition, we are the first
to evaluate whether LDR remain clinically stable or tend to
decline and whether risk factors for clinical and functional
deterioration could be prospectively identified.
of ADL dependency. Eight ADLs are measured in the RAI/
MDS on a scale of 0 to 4 (0 ϭ no dependence; 4 ϭ com-
plete dependence): eating, dressing, bathing, toileting, per-
sonal hygiene, bed mobility, transfer, and locomotion. The
interrater reliabilities for the ADL measures used in this
scale ranged from 0.89 to 0.98 (21). The additive scale
therefore had a range of 0 to 32 (0 ϭ no dependency; 32 ϭ
complete dependency in all ADLs). The mean ADL score
for all residents was 18 Ϯ 11, and the median was 19. The
cut-off score for the lowest quartile, our LDR, was 8.
We compared our ADL scale with a published ADL scale
developed from RAI variables—the RUG-III ADL Index
scale, which succinctly summarizes the effects of resident
functionality (22). The measure of physical function for the
multivariate analysis was the mean value for a composite of
ADL Index that combined self-performance in “late-loss
ADLs”: bed mobility, bed–chair transfer, eating, and toilet-
ing. The range for this scale in our data set is 3 to 15, and
the LDR had a mean score of 3.1 Ϯ 0.68 versus residents
with medium and high ADL dependency (M/HDR) who
had a mean score of 10.3 Ϯ 3.3 ( p Ͻ .001).
METHODS
Type of Study and Data
We conducted a cross-sectional, six-month follow-up
study using secondary data analysis of a sample of nursing
home residents from ten states. Data was obtained from the
Resident Assessment Instrument (RAI) Evaluation Study,
which sampled a different cohort of residents in the same
2
54 nursing homes in 10 states in 1990 and 1993. The RAI
is the resident assessment system mandated in 1991 by the
Health Care Finance Administration (HCFA), which in-
cludes the Minimum Data-Set (MDS), an assessment in-
strument that evaluates residents’ cognitive, behavioral,
functional, and medical status. HCFA sponsored the RAI
evaluation study to assess the RAI’s impact on nursing home
care. A detailed account of the sampling strategy, methodol-
ogy, and results of RAI evaluation study are beyond the
scope of this paper and are published elsewhere (20).
Multi-stage sampling was used within the ten states cho-
sen for the study. Data for the RAI Evaluation Study were
collected in four waves (two pre- and two postimplementa-
tion) by research nurses in each of the participating facili-
ties. Waves 1 (fall, 1990) and 3 (spring 1993) had pre- and
postimplementation baseline data, and waves 2 (spring,
All the independent variables used were taken from the
MDS (research nurses collected this data following standard
published directions) (23). Variables included demographic
and social variables; advance directives and legal oversight;
chronic disease diagnoses; diagnoses of dementia, depres-
sion, and anxiety; cognitive performance; gait; balance;
nutritional status; behaviors suggesting delirium and de-
pression; behavioral problems; medications; and physical
restraints. The average interrater reliability of MDS vari-
ables chosen for analysis ranged from 0.5 to 0.98 (21).
Most variables studied were measured by a single MDS
question. However, the cognitive performance and the de-
pressed behavior variables were composites of several ques-
tions (24,25). The MDS Cognitive Performance Scale (CPS)
has been validated against the Mini Mental Status Examina-
tion and the Test for Severe Impairment (24). The CPS is a
seven-category rating scale (range 0–6), and the items used
for the CPS have an average interrater reliability of 0.85
(24). The MDS Sad or Anxious Mood Scale is based on 12
MDS items and exhibits moderate to good specificity and
sensitivity when compared with clinical judgments made by
trained facility staff (25).
1
991) and 4 (fall, 1993) had six-month follow-up data. The
states were chosen to reflect different reimbursement strate-
gies and staffing and included California, Connecticut,
Iowa, Maryland, Minnesota, Ohio, Oregon, Tennessee,
Texas, and Virginia. The facilities in each state included ur-
ban and rural nursing homes. Ninety-five percent of the fa-
cilities were retained throughout all four waves. In each fa-
cility, an average of eight residents was randomly sampled.
The sample of residents chosen, although not a population-
based sample of all U.S. nursing home residents, was a
probability sample chosen specifically to represent the gen-
eral nursing home population in ten key states. Within the
chosen sampling frame, weights defining each sampled res-
ident’s probability of selection were available.
Model Development and Statistical Analysis
The LDR were compared with all others using standard
descriptive statistics including the t test for continuous vari-
ables and the chi-square test of association for categorical
variables. Variables analyzed were chosen on the basis of
clinical experience and previous literature and represented
important characteristics associated with ADL disability:
sociodemographics, advance directives; cognitive perfor-
mance; depressive symptoms and behaviors; nutritional sta-
tus; sensory problems; medications; and chronic medical,
neurological, and psychiatric diseases. Variables signifi-
cantly associated with low-dependency status were then en-
tered into a logistic regression model to assess their multi-
variate association.
For this study, we pooled data from the 1990 and 1993
cohorts to construct a synthetic cohort of residents with
baseline and 6-month follow-up data. We excluded resi-
dents with length of stay less than 60 days and age less than
6
5 years (study N ϭ 3955).
Variables and Their Measurement
We defined low ADL-dependent residents (LDR, n ϭ
9
85) as those with the fewest ADL limitations, determined
For our longitudinal analysis, we first compared the dif-
ferent six-month outcomes of the LDR and M/HDR with re-
by their presence in the lowest quartile of an additive scale