Validation of a Fall Risk Index in Stroke Rehabilitation

E. Olsson, BSc,* B. Löfgren, PhD,† Y. Gustafson, MD,‡ and L. Nyberg, PhD§

A fall risk index has previously been developed to identify fall-prone individuals in
stroke rehabilitation. The purpose of this study was to validate the predictive accuracy
of the index. The validation sample (n ⫽ 158) consisted of patients admitted to a
specialized geriatric stroke rehabilitation ward. The index was scored for each subject,
and the relationship between the score and falls was assessed. The index was then
remodeled and cross-validated in the sample from which it was derived (model fit
sample, n ⫽ 135). The total index score (0-11) was significantly connected with the time
to first fall (hazard ratio, 1.22; confidence interval [CI], 1.03-1.44). However, the
classification of subjects into groups with low, intermediate, and high risk of falling
could not be correlated with the time to first fall. A remodeled index contained 3 of the
separate original index items. Its relationship to fall risk in the model fit sample was
(hazard ratio, 1.82; CI, 1.38-2.40). The fall risk index showed some correlation with the
fall risk among patients in stroke rehabilitation, but the results indicate that it should
be modified to reach acceptable accuracy. A remodeled index showed a higher
association with fall risk. Key Words: Accidental falls— cerebrovascular disease—
geriatrics—prediction—rehabilitation.
© 2005 by National Stroke Association

Falls are one of the most frequent complications among
patients with stroke during rehabilitation and hospital
care.1,2 The frequency of falls in the geriatric stroke reha-
bilitation setting is 580 per 100 person-years.3 A study of
medical complications after stroke established that 25%
of the patients suffered from falls during hospital admis-
sion.1
It can be assumed that injuries and other consequences
of falls, such as restricted activity as a result of the fear of
From the *Department of Community Medicine and Rehabilita-
tion, *Physiotherapy, †Occupational Therapy, and ‡Geriatric Medi-
cine, University of Umeå, Umeå, Sweden; and §Department of
Health Sciences, Luleå University of Technology, Boden, Sweden.
Received September 20, 2004; accepted November 5, 2004.
Supported by the Swedish Research Council (ref no. 2002-27-VX-
14172), the Vardal Foundation (ref no. V98 085), Umeå University
Foundation of Medical Research, the 1987 Foundation for Stroke
Research, The Swedish Society of Medicine (ref no. 340.0), and the
Kempe Foundations.
Address reprint requests to E. Olsson, PT, BSc, Department of
Community Medicine and Rehabilitation, Physiotherapy, Umeå Uni-
versity, SE-901 87 Umeå, Sweden. E-mail: eva.olsson@physiother.
umu.se.
1052-3057/$—see front matter
© 2005 by National Stroke Association
doi:10.1016/j.jstrokecerebrovasdis.2004.11.001
new falls,4 are likely to have a negative effect on the
rehabilitation process. Furthermore, hospital falls among
stroke patients have been found to be a significant pre-
dictor of falls after discharge from the hospital, which in
turn were associated with lower activity levels, depressed
mood, and pressure on carers.5 The risk of falls has been
shown to be twice as high,6 and the risk of fractures 2-4
times as high,7 among long-term stroke survivors than
among control subjects.
Thus the identification of fall-prone stroke patients is of
great importance. Studies have been carried out with the
aim of identifying risk factors predictive of falls, and a
number of such factors have been suggested. Postural
sway,8 the rate of rise in force in sit-to-stand move-
ment,9,10 increased motor response time to visual stimuli,11
rightward-orienting bias among right hemisphere stroke
patients,12 and depressive symptomatology6,13 have been
associated with increased fall risk. It has also been shown
that patients who experienced at least one fall had signifi-
cantly lower functional independence, balance, and arm,
foot, and leg motor scores compared with non-fallers.14
Among community-dwelling stroke survivors, 1 study re-
ported that repeat fallers had significantly more reduced
arm function and activities of daily living (ADL) ability than
those who did not report any falls,15 and another study

Journal of Stroke and Cerebrovascular Diseases, Vol. 14, No. 1 (January-February), 2005: pp 23–28 23
24 E. OLSSON ET AL.

Table 1. Fall risk index for patients in stroke rehabilitation* The validation sample in this study comprised 158 pa-
tients seen from 1997 to 1998. Like the model fit sample,
Factors Score the validation sample comprised a 1-year sample of
stroke patients consecutively admitted to a specialized
Sex Female 0
geriatric stroke rehabilitation ward at Umeå University
Male 2
Hospital, Sweden. The inclusion criteria were the same
Katz ADL score23 A-D 0
E-G 2 for both samples; the patients had cerebrovascular acci-
Urinary continence, Katz Independent 0 dents or other clinically similar conditions and needed
ADL categorization Dependent 2 further rehabilitation after the acute phase. They were
(incontinent) transferred from acute care clinics usually 2-4 weeks after
Postural stability score22 ⱖ10/14 0 the stroke.
⬍10/14 1 All patients who met the inclusion criteria were in-
Signs of motor None or unilateral 0 cluded after informed consent had been obtained, either
impairment Bilateral 1 from the patients themselves or for a few subjects from
Signs of visuospatial No 0 their relatives. Nine patients declined to participate.
hemi-neglect Yes 1
Three patients, who were completely immobile and bed-
Bilateral brain lesions, No 0
ridden throughout their entire stay, were subsequently
including leukoariosis Yes 1
Use of diuretics, No 0 excluded from analysis because they were judged to be
antidepressants, or Yes 1 not at risk for falling and because rehabilitation was not
sedatives possible. Thus 158 patients remained. Their basic charac-
teristics are summarized in Table 2.
Index values range from 0 to 11: 0-4 indicates a minor fall risk; Table 2 gives comparisons between the 2 samples. The
5-7, an intermediate risk; and 8-11, a high risk. samples differ from one another in that the patients from
*Data from Nyberg and Gustafson (1997).18
the new sample turned out to be more ADL-dependent

showed that accumulated deficits such as motor and sen-

sory impairments increased fall risk, whereas accumulated Table 2. Study sample characteristics at inclusion in study
motor, sensory, and visual impairments seemed to reduce
1997–1998 1991–1992*
the risk of falling.16 Among community-dwelling women
n ⫽ 158 n ⫽ 135
with stroke onset more than 1 year earlier, self-perceived
difficulties with balance while dressing and self-perceived Mean ⫾ SD age (years) 76.4 ⫾ 8.6 74.8 ⫾ 8.9
residual balance problems proved to be independent risk Sex (n; M/F) 72/86 69/66
factors for falls.17 Diagnoses (n)
A fall risk index was previously developed to identify Nonembolic cerebral 93 74
fall-prone individuals in stroke rehabilitation.18 Based on infarctions
a multiple Cox regression model, this index summarizes Embolic cerebral infarctions 31 18
the presence of 8 risk indicators (Table 1). Unspecified cerebral 3 10
infarctions
To evaluate the usefulness and accuracy of the predic-
Cerebral hemorrhages 27 19
tion that any regression model gives, cross-validation
Subarachnoidal/subdural 4 5
studies should be applied to a second sample. Predictor hemorrhage
variables found to be important for 1 sample would not Brain tumors (operated on) — 4
necessarily be accurate for another sample,19 owing to Traumatic brain injuries — 3
chance variations in the correlations between variables. Brain abscesses (operated on) — 2
The sample selection and sample size may also influence Postural instability 75,9% 71,6%
which variables will be selected for the model. The pur- Bilateral motor impairment 22,2% 71,9%
pose of this study was therefore to validate the predictive Bilateral brain lesions 65,2% 44,0%
accuracy of the fall risk index using a new sample of Urinary incontinence 67,7% 61,5%
stroke patients. Visuospatial hemi-inattention 57,0% 46,7%
Risk medicine 65,8% 52,0%
Median Katz ADL scores (first F (E, G) F (C, F)
Materials and Methods and third quartiles)23
Median MMSE scores (first and 16 (0, 25) 20 (7, 27)
Patients third quartiles)22
The model fit sample, from which the Cox regression †Katz et al, (1963).23
model was derived, consisted of 135 patients included ‡Folstein et al, (1975).22
from 1991 to 1992 and has been described previously.18 *Data from Nyberg and Gustafson (1997).18
VALIDATION OF A FALL RISK INDEX IN STROKE REHABILITATION 25

Table 3. Single Cox regression analyses of the associations between the separate index items and time to first fall
within 8 weeks

97/98 97/98 91/92 91/92

Variable Hazard ratio P value Hazard ratio P value

Male sex 1.65 (0.9–3.1) .118 2.08 (1.2–3.7) .013

Katz ADL score E-G 2.56 (0.8–8.4) .123 6.43 (2.0–20.7) .002
Urinary incontinence 1.56 (0.7–3.6) .295 4.05 (1.7–9.5) .001
Postural stability score ⬍10/14 4.50 (1.1–18.7) .039 3.85 (1.4–10.7) .010
Signs of bilateral motor impairment 1.14 (0.6–2.3) .723 3.64 (1.3–10.1) .013
Signs of visuospatial hemi-inattention 2.57 (1.2–5.4) .013 2.16 (1.2–3.9) .010
Bilateral brain lesions, including leukoariosis 0.54 (0.3–0.5) .053 2.03 (1.2–3.6) .014
Use of diuretics, antidepressants, or sedatives 1.11 (0.6–2.2) .772 1.62 (0.9–2.9) .100

and somewhat more cognitively impaired than those
from the old sample.
In accordance with the model fit study, the study
period was set to 8 weeks (56 days) from admission to the
geriatric rehabilitation unit up to discharge or death. The
patients were thus studied for a median period of 34.5
days, (range, 3 to 56 days).
inattention was defined as a positive finding in either of
these 2 ways.
The fall risk index was then scored for each subject, as
indicated in Table 1 All staff members involved in the
care of the patients were blinded to the score.
Falls

Assessments Falls were defined as incidents where the subject, due

Patient characteristic data were collected from the pa-
tients themselves or their relatives, as well as from med-
ical records during the first week of hospitalization at the
geriatric rehabilitation unit. Computed tomography of
the head was done on all subjects, providing information
on the size, nature, and localization of the brain injury.
Assessments were made using the following instru-
ments:
1. The mini mental state examination (MMSE),20 to
assess cognitive state. The score ranges from 0 to 30, with
ⱕ23 points indicating significant cognitive impairment.
2. The motricity index (MI),21 to assess upper and
lower limb motor function. The maximum score for each
body side is 100. In this study the mean scores for both
body sides were noted. In the fall risk index, it was noted
whether or not bilateral motor impairment was observed.
3 The postural stability subscore of the Brunnström–
Fugl–Meyer Scale.22 This subscore ranges from 0 to 14,
with higher scores indicating less impairment.
4 The Katz ADL index,23 which includes six items:
bathing, dressing, toileting, transfer, continence, and
feeding. The score ranges from A to G, with A indicating
independence in all 6 activities and G indicating total
dependence (O denotes unclassifiable cases).
5 The line bisection test for visuospatial hemi-inatten-
tion,24 where the score ranges from 0 to 9 and ⱕ7 indi-
cates impairment. The presence of perceptual impair-
ment was also estimated in a multidisciplinary team
consensus after a comprehensive clinical assessment, in-
cluding observation during activities. Visuospatial hemi-
to an unexpected loss of balance, came to rest on the floor
or an object below knee height. All such incidents that
occurred during the study period and that came to the
knowledge of the nursing staff were reported on special
fall report forms.
Analyses
Falls and injuries were described as numbers, percent-
ages, and incidence rates with 95% confidence intervals
CIs. The incidence rate of falls was calculated as I ⫽ A/R,
where I represents the incidence, A is the number of falls,
and R is the risk period, that is, the sum of each subject’s
number of days on the ward during the study period. The
rate is presented as number of falls per 100 patient years.
Because of the wide range of the observation periods,
survival analyses were used in describing the fall risk as
a function of time. The association between index score
and the time to first fall was analyzed by a Cox regres-
sion.25 Time to first fall was compared between subjects
belonging to low-, intermediate-, and high-risk groups by
Kaplan–Meier analysis with a log-rank test for statistical
significance.25 Each of the fall risk index items was asso-
ciated with the time to first fall using Cox regression.
On the bases of findings from the foregoing analyses,
remodeling of the index was performed. A principal
components analysis (varimax rotation, eigenvalue ⱖ1)
was used to identify index items closely related to each
other. In cases where 2 or more index items strongly
loaded on the same factor (factor loading ⱖ0.6), the item
that exhibited the strongest bivariate association with the
26
risk of fall (Table 3) was chosen for further remodeling.
Factors with P values ⬎.15 in bivariate analyses (Table 3)
were excluded at this step from further analyses. Step-
wise multifactorial Cox regression modeling was used to
identify the combination of index items showing the best
association with falling in the validation sample (1997-
1998). The remodeled index was then tested in the model
fit sample (1991-1992) using Cox regression and Kaplan–
Meier analyses with the log-rank test.
A P value ⱕ .05 was taken to indicate statistical signif-
icance. The SPSS program package, version 10.0, was
used for data analysis.
Ethics
The study was approved by the Research Ethical Com-
mittee of the Medical Faculty of Umeå University (see,
84/97, dnr 97-70).
Results
Falls and Risk Index Scoring
During the first 8 weeks of rehabilitation, 39 (25%)
patients fell at least once. The incidence rate was 349 (95%
CI, 254-444) falls per 100 patient years, compared with the
corresponding rate of 756 (95 CI%, 616-897) in the model
fit sample. For 5 subjects, data were missing on 1 or more
items, and thus these patients could not be included in
the following analyses.
The median index score (and interquartile range) was 7
(5-8). The score was significantly associated with the time
to first fall (hazard ratio, 1.22; CI, 1.03-1.44). Of those
patients who fell, 1 belonged to the group classified as
having a low risk of falls, which gives a sensitivity of
97%. The specificity was 26%, because 30 of the 114
patients without falls were categorized as being at low
risk.
Figure 1 presents a Kaplan–Meier analysis of fall risk
for subjects in the 3 different fall risk groups. No signif-
icant difference was seen between groups (P ⫽ .07), and
the intermediate- and high-risk groups showed an over-
lapping risk of falls as a function of time. As Table 3
shows, the separate index items low postural score and
visuospatial hemi-neglect had significant associations
with fall risk. Low Katz ADL scores and male sex were
not separately associated with fall risk, but showed
trends toward positive associations. Urinary inconti-
nence, bilateral motor impairment, and use of diuretics,
antidepressants, or sedatives had far from significant
association values. Bilateral brain lesion was very close to
significance, however, with a trend toward a protective
effect.
Remodeling of the Index
The principal component analysis revealed a pattern in
which the index items Katz ADL score E-G, urinary
E. OLSSON ET AL.
Figure 1. Kaplan-Meier analysis of fall risk in subjects assigned to the
low-risk (dotted line), intermediate-risk (solid line), and high-risk (fat solid
line) groups. Log-rank test for validation sample, 5.19, (P ⫽ .075).
incontinence, and postural stability score ⬍10/14 loaded
on a first factor; male sex and use of diuretics, antide-
pressants, or sedatives loaded on a second factor; and
signs of visuospatial hemi-inattention loaded on a third
factor. Bilateral brain lesions and bilateral motor impair-
ment did not load significantly on any of the factors. The
stepwise Cox regression modeling resulted in the follow-
ing items: postural stability score ⬍10/14, signs of visuo-
spatial hemi-inattention, and male sex. Bilateral brain
lesions did not significantly contribute to the association
with fall risk when tried with the other variables and
therefore was excluded.
The 3 remaining items were grouped together to form
an index following an accumulated model. A postural
stability score ⬍10/14 (item 1) gives a score of 1; item 1 ⫹
visuospatial hemi-inattention (item 2) gives a score of 2;
and item 1 ⫹ item 2 ⫹ male sex (item 3) renders a score
of 3. The score was significantly associated with fall risk
in both samples (Fig 2); the hazard ratio (95% CI) was 1.9
(1.4-2.7) in the validation sample and 1.8 (1.4-2.4) in the
model fit sample.
Discussion
This cross-validation study established that the previ-
ously published fall risk index could not be considered
sufficiently accurate. The score was significantly associ-
ated with fall risk, but the suggested labeling of subjects
as low-, intermediate-, and high-risk groups could not be
validated. Further confirmation of insufficient accuracy
was that only 2 of the separate items proved to be
significantly associated with the fall risk when they were
tried in single Cox regression analyses.
In the validation sample there were far fewer falls than
in the previous study, which should affect the outcome of
VALIDATION OF A FALL RISK INDEX IN STROKE REHABILITATION
Figure 2. Kaplan-Meier analyses of the relation of the revised index to the
fall risk among patients in the validation sample (A) and in the model fit
sample (B). The dotted line indicates a score of 0; the solid line, a score of 1;
the fat dotted line, a score of 2; and the fat solid line, a score of 3. Log-rank
test for the validation sample, 18.2 (P ⬍ .001), for the model fit sample, 23.3
(P ⬍ .001).
the index. The remarkable decrease in the number of falls
(56%) might be due to the changes in the patients’ envi-
ronment and a greater fall risk attention among staff since
the previous study. The geriatric clinic moved to newly
built premises more suitable for rehabilitation. At the
same time, the rehabilitation staff grew, particularly oc-
cupational therapists and physiotherapists. In contrast,
however, there was a greater demand to reduce lengths
of stay due to the increasing number of patients waiting
for geriatric stroke rehabilitation.
27
The insufficient accuracy of the fall risk index may be
due to various causes. The low fall rates may have
produced less statistical power in the analyses. In addi-
tion, many of the patients who did not fall in this study
might have been fallers in the previous study. There were
also small differences between the samples’ characteris-
tics, with the patients in the later sample being somewhat
older, more ADL-dependent, and more cognitively im-
paired.
During the index analysis, it came to our attention,
somewhat surprisingly, that the presence of bilateral
brain lesions, including leukoariosis, showed a trend
toward a strong protective effect. One possible explana-
tion for this finding is that some of these patients were so
disabled that they had not been physically active. More-
over, the way in which the diagnoses were determined
may have changed over the years since the index was
created.
This study indicates a predictive accuracy at least on
the same level as other external validations of fall risk
indexes. A validation of the Downton index among
stroke patients in geriatric rehabilitation showed a sensi-
tivity of 91% and a specificity of 27%.26 In an external
validation of the Tinetti fall risk index among communi-
ty-dwelling people, the most accurate cutoff score found
resulted in a sensitivity of 70% and a specificity of 52%.27
A low specificity of the index is more acceptable than a
low sensitivity, but a higher specificity would be of value
for a more accurate prediction.
The remodeled index showed a better accuracy in both
samples, and its reduction to 3 items should make it more
user-friendly. The new model has also been constructed
according to an accumulated model that was earlier
proven to be productive.16 This accumulated model dis-
tinguishes between those suffering only from impaired
postural stability and those with additional impairments,
such as visuospatial hemi-inattention. In this model, male
sex further contributes to fall risk when the other factors
are present. Contrary to what has been found in commu-
nity populations of elderly persons,28 a few studies indi-
cate that among hospitalized elderly patients, men are
more likely to fall than women.29
Other risk factors, including impaired cognition, de-
pressive symptomatology,6,13 motor performance (in,
e.g., sit-to-stand movement),9,10 and perceived balance
disturbances and perceived difficulties with balance
while performing common activities such as dressing,17
have not been taken into account in this article because
this is a validation study of a previously published in-
strument. However, these are important factors that
merit more careful examination in future research.
The correct classification of patients with different lev-
els of fall risk is essential to high-risk prevention strate-
gies. Such a strategy was successfully tried in frail elderly
patients in residential care,30 and the remodeled index
28
may be of value in preventive trials directed toward
patients who have sustained stroke.
In conclusion, when externally validated, the fall risk
index showed some relationship to fall risk among pa-
tients in a stroke rehabilitation setting, but the results
indicate that this index should be modified to achieve
acceptable accuracy. The remodeled index showed a
greater association with fall risk and could well be used
as a foundation for future research in this area.
References
1. Langhorne P, Stott DJ, Robertson L, et al. Medical com-
plications after stroke: a multicenter study. Stroke 2000;
31:1223-1229.
2. Dromerick A, Reding M. Medical and neurological com-
plications during inpatient stroke rehabilitation. Stroke
1994;25:358-361.
3. Nyberg L, Gustafson Y. Patient falls in stroke rehabilita-
tion, challenge to rehabilitation strategies. Stroke 1995;
26:838-842.
4. Vellas B, Cayla F, Bocquet H, et al. Prospective study of
restriction of activity in old people after falls. Age Age-
ing 1987;16:189-193.
5. Forster A, Young J. Incidence and consequences of falls
due to stroke: a systematic inquiry. BMJ 1995;311:83-86.
6. Jorgensen L, Engstad T, Jacobsen BK. Higher incidence
of falls in long-term stroke survivors than in population
controls: depressive symptoms predict falls after stroke.
Stroke 2002;33:542-547.
7. Ramnemark A, Nyberg L, Borssen B, et al. Fractures after
stroke. Osteoporos Int 1998;8:92-95.
8. Sackley CM. Falls, sway, and symmetry of weight-bear-
ing after stroke. Int Disabil Stud 1991;13:1-4.
9. Cheng PT, Liaw MY, Wong MK, et al. The sit-to-stand
movement in stroke patients and its correlation with
falling. Arch Phys Med Rehabil 1998;79:1043-1046.
10. Cheng PT, Wu SH, Liaw MY, et al. Symmetrical body-
weight distribution training in stroke patients and its
effect on fall prevention. Arch Phys Med Rehabil 2001;
82:1650-1654.
11. Mayo NE, Korner-Bitensky N, Kaizer F. Relationship
between response time and falls among stroke patients
undergoing physical rehabilitation. Int J Rehabil Res
1990;13:47-55.
12. Webster JS, Roades LA, Morrill B, et al. Rightward ori-
enting bias, wheelchair maneuvering, and fall risk. Arch
Phys Med Rehabil 1995;76:924-928.
13. Ugur C, Gucuyener D, Uzuner N, et al. Characteristics of
falling in patients with stroke. J Neurol Neurosurg Psy-
chiatry 2000;69:649-651.
E. OLSSON ET AL.
14. Teasell R, McRae M, Foley N, et al. The incidence and
consequences of falls in stroke patients during inpatient
rehabilitation: factors associated with high risk. Arch
Phys Med Rehabil 2002;83:329-333.
15. Hyndman D, Ashburn A, Stack E. Fall events among
people with stroke living in the community: circum-
stances of falls and characteristics of fallers. Arch Phys
Med Rehabil 2002;83:165-170.
16. Yates JS, Lai SM, Duncan PW, et al. Falls in community-
dwelling stroke survivors: an accumulated impairments
model. J Rehabil Res Dev 2002;39:385-394.
17. Lamb SE, Ferrucci L, Volapto S, et al. Risk factors for
falling in home-dwelling older women with stroke: the
Women’s Health and Aging Study. Stroke 2003;34:494-
501.
18. Nyberg L, Gustafson Y. Fall prediction index for patients
in stroke rehabilitation. Stroke 1997;28:716-721.
19. Jongbloed L. Prediction of function after stroke: a critical
review. Stroke 1986;17:765-776.
20. Folstein MF, Folstein SE, McHugh PR. “Mini-mental
state”: a practical method for grading the cognitive state
of patients for the clinician. J Psychiatr Res 1975;12:189-
198.
21. Collin C, Wade D. Assessing motor impairment after
stroke: a pilot reliability study. J Neurol Neurosurg Psy-
chiatry 1990;53:576-579.
22. Fugl-Meyer AR, Jaasko L, Leyman I, et al. The post-
stroke hemiplegic patient, 1. A method for evaluation of
physical performance. Scand J Rehabil Med 1975;7:13-31.
23. Katz S FA, Moskowitz RW, Jackson BA, et al. Studies of
illness in the aged: the index of ADL. A standardized
measure of biological and psychosocial function. JAMA
1963:914-919.
24. Friedman PJ. Spatial neglect in acute stroke: the line
bisection test. Scand J Rehabil Med 1990;22:101-106.
25. Dawson-Saunders B, Trapp RG. Basic & Clinical Biosta-
tistics. 2nd ed. East Norwalk, CT: Appleton & Lange,
1994.
26. Nyberg L, Gustafson Y. Using the Downton index to
predict those prone to falls in stroke rehabilitation.
Stroke 1996;27:1821-1824.
27. Raiche M, Hebert R, Prince F, et al. Screening older
adults at risk of falling with the Tinetti balance scale.
Lancet 2000;356:1001-1002.
28. Campbell AJ, Spears GF, Borrie MJ. Examination by
logistic regression modelling of the variables which in-
crease the relative risk of elderly women falling com-
pared to elderly men. J Clin Epidemiol 1990;43:1415-
1420.
29. Nyberg L, Gustafson Y, Janson A, et al. Incidence of falls
in three different types of geriatric care: a Swedish pro-
spective study. Scand J Soc Med 1997;25:8-13.
30. Jensen J, Lundin-Olsson L, Nyberg L, et al. Fall and
injury prevention in older people living in residential
care facilities: a cluster randomized trial. Ann Intern
Med 2002;136:733-741.