Development of an assessment sheet for fall prediction in stroke inpatients in convalescent rehabilitation wards in Japan
Environ Health Prev Med
Development of an assessment sheet for fall prediction in stroke inpatients in convalescent rehabilitation wards in Japan
Youichi Nakagawa 0 1 2 3 4 5 6 7
Katsuhiko Sannomiya 0 1 2 3 4 5 6 7
Makiko Kinoshita 0 1 2 3 4 5 6 7
Tsutomu Shiomi 0 1 2 3 4 5 6 7
Kouhei Okada 0 1 2 3 4 5 6 7
Hisayo Yokoyama 0 1 2 3 4 5 6 7
Yukiko Sawaguti 0 1 2 3 4 5 6 7
Keiko Minamoto 0 1 2 3 4 5 6 7
Chang-nian Wei 0 1 2 3 4 5 6 7
Shoko Ohmori 0 1 2 3 4 5 6 7
Susumu Watanabe 0 1 2 3 4 5 6 7
Koichi Harada 0 1 2 3 4 5 6 7
Atsushi Ueda 0 1 2 3 4 5 6 7
0 M. Kinoshita Hatsudai Rehabilitation Hospital , Tokyo , Japan
1 K. Sannomiya S. Watanabe Kumamoto Kinoh Hospital , Kumamoto , Japan
2 Y. Nakagawa (&) K. Minamoto C. Wei S. Ohmori A. Ueda Department of Preventive and Environmental Medicine, Graduate School of Medical and Pharmaceutical Sciences, Kumamoto University , 1-1-1 Honjo, Kumamoto 860-8556 , Japan
3 K. Harada Department of Microbiology and Environmental Chemistry, School of Health Sciences, Kumamoto University , Kumamoto , Japan
4 Y. Sawaguti Ohta Atami Hospital , Fukushima , Japan
5 H. Yokoyama Ukai Rehabilitation Hospital , Nagoya , Japan
6 K. Okada Chikamori Rehabilitation Hospital , Kochi , Japan
7 T. Shiomi Morinomiya Hospital , Osaka , Japan
Objective We conducted a study to develop an assessment sheet for fall prediction in stroke inpatients that is handy and reliable to help ward staff to devise a fall prevention strategy for each inpatient immediately upon admission. Methods The study consisted of three steps: (1) developing a data sampling form to record variables related to risk of falls in stroke inpatients and conducting a follow-up
Fall; Stroke; Assessment; Risk factor; Rehabilitation
survey for stroke inpatients from their admission to
discharge by using the form; (2) carrying out analyses of
characteristics of the present subjects and selecting
variables showing a high hazard ratio (HR) for falls using the
Cox regression analysis; (3) developing an assessment
sheet for fall prediction involving variables giving the
integral coefficient for each variable in accordance with the
HR determined in the second step.
Results and discussion (1) Subjects of the present survey
were 704 inpatients from 17 hospitals including 270 fallers.
(2) We selected seven variables as predictors of the first
fall: central paralysis, history of previous falls, use of
psychotropic medicines, visual impairment, urinary
incontinence, mode of locomotion and cognitive
impairment. (3) We made 960 trial models in combination with
possible coefficients for each variable, and among them we
finally selected the most suitable model giving coefficient
number 1 to each variable except mode of locomotion,
which was given 1 or 2. The area under the ROC curve of
the selected model was 0.73, and sensitivity and specificity
were 0.70 and 0.69, respectively (4/5 at the cut-off point).
Scores calculated from the assessment sheets of the present
subjects by adding coefficients of each variable showed
normal distribution and a significantly higher mean score in
fallers (4.94 ± 1.29) than in non-fallers (3.65 ± 1.58)
(P = 0.001). The value of the Barthel Index as the index of
ADL of each subject was indicated to be in proportion to
the assessment score of each subject.
Conclusion We developed an assessment sheet for fall
prediction in stroke inpatients that was shown to be
available and valid to screen inpatients with risk of falls
immediately upon admission.
Stroke patients are encouraged to start rehabilitation in
the early stage from the onset of stroke because early
mobilization and more aggressive rehabilitation are
recognized to improve functional outcomes of the patients
]. From this view point, the application of an
effective stroke management system in combination with
rehabilitative treatments from the acute phase of stroke is
Usually in Japan, stroke patients undergo rehabilitation
in three stages: acute, convalescent and maintenance. It is
necessary to build a systematized cooperation between
hospitals for those three stages so that stroke inpatients can
receive the appropriate rehabilitation smoothly at each
The purpose of convalescent rehabilitation is to regain
mobility and independence in activity of daily life (ADL)
and to prevent muscle weakness. It has been proven that
aggressive rehabilitation from admission to convalescent
rehabilitation wards after receiving treatment during the
acute period may rapidly improve the ADL of stroke
]. However, inpatients are not accustomed to the
ward environment for a while after admission and that
situation may increase the chance of falls in the early stage
of rehabilitation [
]. In addition, stroke inpatients often
have cognitive impairment and high cortical dysfunction,
increasing the risk of falls further [
Falls may have serious consequences for patients
receiving rehabilitation, such as injury and disability [
Moreover, the psychological changes such as fear of falls
may result in self-induced restriction in each activity [
]. Those consequences have a negative effect on the
rehabilitation process and its outcome and may result in
extended hospitalization and increased medical expense
The risk factors for falls are based on multiple factors
such as the physical faculty, neuropsychological and
environmental situations of inpatients. It is necessary to
consider external and internal factors inclusively to
develop fall prevention strategies for inpatients undergoing
]. On the other hand, unnecessary
restriction of activity to prevent falls goes against the
purpose of rehabilitation. Therefore, the staff in the
rehabilitation wards faces a dilemma balancing aggressive
rehabilitation and fall prevention.
In convalescent rehabilitation wards the staff of various
occupational categories cooperates to support rehabilitation
of stroke inpatients, and all of them commonly recognize
the risk of falls among inpatients [
]. It is also
recognized that falls frequently occur within the first week
after starting rehabilitation .
To date, there have been many reports on the method of
assessment for fall prediction among inpatients in
rehabilitation wards [
]. However, there are hardly any reports
on the method focusing on stroke inpatients who are the
majority of patients in convalescent rehabilitation wards in
Japan. These facts indicate that it is necessary for the staff
in convalescent rehabilitation wards to provide a handy and
reliable method of screening inpatients who will be prone
to falls immediately upon admission.
From the above viewpoint we developed a form of
assessment sheet for fall prediction in stroke patients so
that ward staff can use it to devise a fall prevention strategy
immediately upon admission of each inpatient by using
available and appropriate data that the staff can easily
obtain in their work place.
We developed an assessment sheet for fall prediction for
stroke inpatients by three steps as follows.
Data sampling for selecting variables as an index for an
assessment sheet for fall prediction in stroke inpatients
Development of a data sampling form for recording
factors related to falls of inpatients in convalescent
We organized a workshop team with ten members
consisting of medical and co-medical staff engaged in convalescent
rehabilitation wards and research staff of the departments of
epidemiology and social medicine in Kumamoto University
to develop an assessment sheet for stroke inpatients in
convalescent rehabilitation wards. Firstly, we established
the definition of fall and diagnosis of stroke. A fall was
defined as follows: when part of the body above the knee
comes in contact with the floor surface against the patient’s
]. The diagnosis of stroke was based on brain CT,
MRI and clinical examination by neurologists.
Next, we collected factors related to falls on the basis of
expert knowledge of the workshop members and on the
basis of articles on falls associated with rehabilitation.
Among factors collected we selected items as independent
variables of the present sampling form to be associated
with falls of stroke inpatients in convalescent rehabilitation
wards and to be easy to use for the staff of wards
immediately upon admission of each patient.
The factors chosen as independent variables for falls
were as follows: age, sex, presence or absence of central
paralysis, history of previous falls from the day of stroke
onset to the day before admission to the convalescent
rehabilitation ward (history of previous falls), use of
psychotropic medicines, consciousness disturbance, delirium,
depression, visual impairment, sensory disturbance, ataxia,
high cortical dysfunctions (e.g., apraxia, aphasia, unilateral
spatial neglect and attention disturbance), urinary
incontinence, fecal incontinence, the mode of locomotion (walk
independently, walk with a cane, walk with a walker, use a
wheelchair and on a stretcher), pain, cognitive impairment
As for determination of cognitive impairment, we
adopted the Revised Hasegawa’s Dementia Scale
(HDSR). The HDS-R measures the level of cognitive impairment
of not only ordinary patients, but also disabled patients
with motor and visual impairment [
]. As for
determination of ADL, we adopted the Barthel Index (BI). BI
measures personal levels of functional independence in
Using the above factors we developed a questionnaire
for data sampling of falls in convalescent rehabilitation
wards. We developed three types of questionnaire for each
patient for use upon admission, at every time a fall
occurred and upon discharge.
Method of questionnaire survey
There were 17 convalescent rehabilitation wards
participating in the present survey. They were all members of the
conference of convalescent rehabilitation wards in Japan.
The structures, procedure of rehabilitation and organization
of staff were similar among these wards.
The present sampling survey was approved by the ethics
committee of each facility. All of the inpatients in the 17
wards, after receiving acute treatment for cerebrovascular
attack, were registered as the subjects of the present study
after obtaining written informed consent. If the inpatients
did not have the ability to consent, we got the consent from
their family. The subjects were followed up from their
admission to discharge. The follow-up period per inpatient
was less than 3 months. The study period was from 1 June
2004 to 31 June 2005.
The occurrence of falls was reported by the staff,
inpatients or inpatients’ families, and all data of the subjects
were confirmed and recorded by the staff of each facility
following the format that we sent. After finishing the
follow-up, the recorded formats were sent to the database
settled on by the working group from each ward.
Methods of analyses for selecting variables
related to falls
First we conducted simple and cross tabulations of the
sampling data. For determination of significance in
differences between fallers and non-fallers, v2-test, t test,
Mann–Whitney U test and Wilcoxon sign rank test were
After cross tabulation we performed the univariate Cox
regression analysis to select preliminary variables to
involve in the assessment sheet for fall prediction using the
items showing significant differences between fallers and
Next, using variables selected by the univariate analysis
indicating P \ 0.10 as the significant value of the hazard
ratio (HR) to falls, we performed the multivariate Cox
regression analysis to select variables as appropriate items
to include in the assessment sheet. The variables finally
selected were indicated to be P \ 0.10 in the significant
value of the HR.
Development of assessment sheet for fall prediction
Using those variables selected by the statistical procedure
as mentioned above, we developed the assessment sheet
for fall prediction according to the procedure as follows.
Firstly, we chose appropriate integral numbers for each
variable as a possible coefficient in accordance with the
value of HR and 95% CI by the multivariate Cox
regression analysis (see Table 2), if characteristics of the
significant variables were present for each subject. If those
were absent for each subject, the number given to it was 0.
Secondly, we made trial models of the assessment sheet in
combination with all of the numbers given to each variable
as a possible coefficient. The total score of each assessment
sheet was calculated by adding the coefficient of each
variable and used as the score of fall risk of each subject.
Thirdly, we performed the receiver operating
characteristics (ROC) analysis and determined the area under the
ROC curve (AUC) to all of the trial models. Finally, among
those models one model indicating the highest value of
AUC with appropriate balance of sensitivity and specificity
was selected as the most suitable assessment sheet.
Statistical analysis All data were analyzed by SPSS ver.11 statistical software programs.
Characteristics of the subjects
A total of 704 stroke patients were admitted to
convalescent rehabilitation wards, and all of them were registered as
the subjects of the present study with the prescribed onset
form. Among them, 270 (38% of total subjects) were
fallers. The range of fall rates in each hospital was 35–40%.
Of 270 fallers, 37% had their first fall within 10 days after
their admission, and also 60% of them fell within 4 weeks
and 46% of them experienced two or more falls, i.e.,
The characteristics of the subjects as fallers and
nonfallers are shown in Table 1. The items indicating
significantly higher rates in fallers than in non-fallers were as
follows: age (P = 0.002), history of previous falls
(P \ 0.001), side of hemiparesis (P \ 0.001), ten clinical
signs (P \ 0.001&P = 0.013), use of psychotropic
medicines (P = 0.004), mode of locomotion (P \ 0.001),
median (Me) of HDS-R (P \ 0.001), days after onset
(P = 0.020) and days of hospitalization (P \ 0.001).
Statistics of BI were as follows (not shown in the table).
The median (Me) BI of the subjects was 55 upon admission
and 80 upon discharge, showing a significant difference
between admission and discharge (P \ 0.001). Non-fallers
(60) showed significantly (P \ 0.001) higher BI than
fallers (40) upon admission. In both non-fallers and fallers the
BI upon discharge (85 for non-fallers and 70 for fallers)
was better than upon admission (P \ 0.001) between
admission and discharge for non-fallers and fallers.
Analyses for selecting the variables related to falls
After cross tabulation we performed univariate and
multivariate regression analyses to select variables to involve in
the assessment sheet using the items of the follow-up study
except showing no significant difference between fallers
and non-fallers. The results of the univariate and
multivariate Cox regression analyses are shown in Table 2. In
the univariate analysis, 14 variables were selected as
preliminary predictors showing strong correlation (P \ 0.10)
to falls, such as presence or absence of central paralysis,
history of previous falls, use of psychotropic medicines,
delirium, visual impairment, sensory disturbance, apraxia,
unilateral spatial neglect, attention disturbance, urinary
incontinence, fecal incontinence, pain, mode of locomotion
and cognitive impairment (the score under 26 in HDS-R).
By the multivariate regression analysis following the
univariate analysis, seven variables with a high HR
(P \ 0.10) were selected as adoptable variables in the
present assessment sheet, such as central paralysis (not
laterality), history of previous falls, use of psychotropic
medicines, visual impairment, urinary incontinence, mode
of locomotion and cognitive impairment (score under 26 in
Development of the assessment sheet for fall
prediction for stroke inpatients
The range of given integral numbers as preliminary
coefficient of each variable transferred values in
accordance with HR and 95% CI by the multivariate Cox
regression analysis was as follows; 1, 2, 3 and 4 for
central paralysis, 1 and 2 for history of previous falls, for
psychotropic medicines, for visual impairment and
for urinary incontinence, 1, 2, 3, 4 and 5 for mode of
locomotion, and 1, 2 and 3 for cognitive impairment
(score under 26 in HDS-R).
By the combination with all possible coefficients
given to seven variables, we developed 960 trial models
of the assessment sheet. For all 960 trial models we
calculated AUC, sensitivity and specificity. As shown in
Table 3, six trial models showed higher AUC value
(0.73) with a smooth curve of ROC than other trial
models. Among six trial models, we chose model 1 (see
Table 3) as the most suitable model because model 1
showed the most reasonable balance of sensitivity and
specificity. As shown in the appendix, the final
coefficient of each variable was as follows: 1 for central
paralysis, for history of previous fall, use of
psychotoropic medicines, visual impairment, urinary incontinence,
score under 26 in HDS-R and 1 for walk with walkers or
2 for using a wheel chair as the mode of locomotion.
Then the total score of the present assessment sheet was
ranged from 0 to 8. When the cut-off point was set at 4
and 5, the sensitivity and specificity were indicated to be
0.70 and 0.69, respectively.
As shown in Table 4, the distribution of the total score
calculated from the present assessment sheet applying data
of the subjects upon admission was proved to be normal by
Kolmogorov–Smirnov test [
] and shows a dose–response
relation curve both in all the subjects and in fallers. The
mean score of fallers (4.95 ± 1.29) was higher than that of
non-fallers (3.65 ± 1.58) (P \ 0.001).
According to the percentile value of the assessment
score calculated from each subject, the present subjects
were classified into three groups. The subjects with scores
of 0 to 2 (0–25th percentile of total subjects) were
classified as risk-1, those with scores of 3 to 4 (25–75th
percentile) as risk-2 and those with scores of 5 to 8 (75–
100th percentile) as risk-3. The rates of falls were 9.2% for
risk-1, 27.4% for risk-2 and 58.3% for risk-3. By Kaplan–
Meier analysis with the log-rank test, a significant
difference was found (log rank statistics, 77.98; P \ 0.001)
among three groups as shown in Fig. 1. The Kaplan–Meier
curve also showed that at least 85% of inpatients in risk-1
underwent rehabilitation without experiencing falls during
the observation period and, in contrast, 20% of inpatients in
risk-3 experienced a fall within 10 days after admission.
As shown in Table 5, the median BIs upon admission
and discharge were 80 and 100 for risk-1, 60 and 80 for
risk-2 and 30 and 60 for isk-3, indicating that the BI of
each subject improved depending on the risk group through
their admission to discharge.
For stroke inpatients in convalescent rehabilitation wards,
falls are the most important adverse event to archive a goal
of each rehabilitation outcome. For staff of rehabilitation
wards receiving such inpatients, they should precisely
predict the risk of falls of each inpatient immediately upon
admission. However, to date, we have no appropriate
method of screening of inpatients who are prone to falls
early after admission. From this view point, we developes
an assessment sheet for fall prediction of stroke inpatients
that can be used effectively by staff of convalescent
rehabilitation wards to design a fall prevention strategy
immediately upon admission of each inpatient using
available data that the staff can easily obtain from each inpatient.
We organized a workshop team on the prevention of
falls in those undergoing rehabilitation consisting of
medical and co-medical staff in convalescent rehabilitation
wards and specialists in preventive medicine of the
Kumamoto University staff.
In the workshop, firstly, we collected items associated
with falls in stroke patients undergoing rehabilitation by
bibliographic references and technical knowledge of
workshop members; among those items collected we
selected items that may be obtained easily by any staff of
the ward immediately upon admission of each inpatient.
Using the items selected we developed case sampling
formats on falls for use upon admission, every time a fall
event occurred and upon discharge, and using those
formats we conducted a follow-up survey for the stroke
inpatients in 17 hospitals with similar ward structures,
rehabilitation procedures and staff organizations. The
procedure of the present study in combination with
qualitative research and quantitative and prospective
research may confirm the validity of the process of data
collecting and of factors related to falls collected from the
In the follow-up study the present subjects showed a fall
rate of 38%, indicating an ordinary fall rate comparable
with previous reports of the rehabilitation wards [
Among 270 fallers 46% of them were recurrent fallers. It
has been indicated that recurrent fallers might show poor
improvement of BI or ADL compared to single and
]. However, we attached importance to the
first fall to be involved in the present assessment sheet
because, as our results clarified, many falls occurred within
a short time after admission and because the previous
report revealed that fallers might develop a psychological
change such as fear of falling resulting in self-induced
restrictions in activity [
The distribution of scores of BI of the present subjects
upon discharge showed significant improvement compared
to those upon admission for both fallers and non-fallers,
indicating that appropriate and effective rehabilitation for
each inpatient has been provided in each rehabilitation
ward. Nevertheless, BI has been recognized as an
important factor related to falls. We excluded BI from the items
of the present assessment sheet because BI was usually
evaluated on the basis of determination of ADL of the
inpatient for about 1 week after admission, and the staff of
rehabilitation wards cannot use the data of BI upon
admission of each inpatient.
We performed univariate and multivariate Cox
regression analysis [
] to select variables to include in the
present assessment sheet from the factors adopted in
the data-sampling format. The Cox regression analysis is
the most suitable method to evaluate the intensity of the
relationship between falls and each co-variable with
common time factors with each other and to determine the
value of coefficient of each variable for falls as the HR.
Those variables selected by the present statistical methods
involve internal factors such as motor, sensory and
cognitive ability and external factors such as drug and devices.
Findings of other studies on falls among inpatients in
rehabilitation wards suggest that such variables selected are
definitely valid for fall prediction in stroke inpatients [
]. These facts indicate the confirmed validity at each
variable selected to be adopted in the present assessment
sheet both clinically and statistically.
By the combination of all possible coefficients to seven
selected variables based on the HR and 95% CI, we
developed 960 trial models and chose 6 models with the
highest value of AUC. Among those six models we
selected model 1 (see Table 3) as the most suitable model
because it indicates an appropriate balance of sensitivity
As shown in the appendix, each of seven variables was
given the coefficient number of the presence of each
corresponding variable as follows: 1 to central paralysis,
history of previous fall, use of psychotoropic medicines,
visual impairment, urinary incontinence, score under in
HDS-R and 1 or 2 to mode of locomotion. Then, the range
of the total score of the present sheet per each subject by
adding coefficient numbers of each variable was 0–8. The
distribution of the scores of each subject based on the
present sheet was shown to be normal. The mean score of
the fallers was significantly higher than that of the
nonfallers (P \ 0.001), and also the dose–response
relationship were shown in the scores of both total subjects and
that of fallers. Those facts indicate that the present sheet is
valid and available for screening stroke inpatients who are
prone to falls.
Therefore, we considered an appropriate use of the
present assessment sheet. According to the percentile value
of each inpatient the subjects were classified into three
groups such as risk-1, risk-2 and risk-3. As shown in Fig. 1,
a significant difference in fall rate by days after admission
was found among three groups by Kaplan–Meier analysis
]. The Kaplan–Meier curve also showed that more than
20% of the inpatients in risk-3 experienced a fall within
10 days after admission and, in contrast, the fallers of the
risk-1 from admission to discharge were only 15%. The
Kaplan–Meier curve indicated that the present assessment
model was proved to be effective for identifying high-risk
inpatients in the early period after admission of
convalescent rehabilitation wards.
As to the BI, it was clarified that the BI of inpatients in
each risk group showed a significant decrease in order of
risk-1 to 3 both upon admission and upon discharge
(Table 5). The result indicates that BI may change
depending on the total score of the present sheet and suggests that it
is not necessary to adopt BI in the assessment sheet.
According with these facts we defined risk-1 to be the
low risk group, risk-2 to be the middle risk group and
risk-3 to be the high risk group for falls, and we advised
the staff of convalescent rehabilitation wards that this
classification should be applied to each inpatient by ward
staff to devise a fall prevention strategy immediately upon
We are now conducting new follow-up research for the
staff of convalescent rehabilitation wards to evaluate the
availability and the validity of using the present assessment
sheet to screen inpatients who are prone to falls and of
conducting the classification of the three risk groups to
develop a strategy for the prevention of falls for each
We developed an assessment sheet consisting of seven
variables for fall prediction for stroke inpatients that can be
applied to screen inpatients with a fall risk immediately
upon admission. The assessment sheet consists of variables
that staff of convalescent rehabilitation wards can easily
obtain from data of inpatients upon admission. We
recommend that the staff of convalescent rehabilitation wards
should use the present assessment sheet immediately upon
admission of each inpatient and classify each inpatient into
three groups according to the assessment score in order to
devise a strategy for fall prevention.
Acknowledgements The study was supported by the Medical Care
Safety Committee of the Liaison Council concerning the National
Convalescent Rehabilitation Care Unit. The authors express many
thanks to Dr. Masahiro Shono, Yuge Hospital, Kumamoto Japan, for
his valuable statistical advice and help.
Final assessment sheet for prediction of falls with the coefficient score
of each variable by presence and absence in each subject
Use of psychotropic medicines
History of previous falls
Mode of locomotion
Walk with walker
1 You choose one item from these
1. Chan KM , Wong SF , Yoong T . Early rehabilitation for stroke patients: a new look . Singapore Med J. 1998 ; 39 : 451 - 5 .
2. Maulden SA , Gassaway J , Horn SD , Smout RJ , Dejong G . Timing of initiation of rehabilitation after stroke . Arch Phys Med Rehabil . 2005 ; 86 : 34 - 40 .
3. Novack TA , Satterfield WT , Lyons K , Kolski G , Hackmeyer L , Connor M. Stroke onset and rehabilitation: time lag as a factor in treatment outcome . Arch Phys Med Rehabil . 1984 ; 65 : 316 - 9 .
4. Hamrin E . Early activation in stroke: does it make a difference? Scand J Rehabil Med . 1982 ; 14 : 101 - 9 .
5. Ishigami S. Rehabilitation approaches to CVA in the acute stage . Nippon Ronen Igakkai Zasshi . 2000 ; 37 : 892 .
6. Paolucci S , Antonucci G , Grasso MG , Morelli D , Troisi E , Coiro P , et al. Early versus delayed inpatient stroke rehabilitation: a matched comparison conducted in Italy . Arch Phys Med Rehabil . 2000 ; 81 : 695 - 700 .
7. Hashimoto Y , Terasaki T , Yonehara T , Tokunaga M , Hirano T , Uchino M. Critical pathway and hospital-hospital cooperation in acute stroke. Reduction of the length of hospital stay . Interventional Neuroradiol . 2000 ; 6 : 251 - 5 .
8. Suzuki T , Sonoda S , Misawa K , Saitoh E , Shimizu Y , Kotake T . Incidence and consequence of falls in inpatient rehabilitation of stroke patients . Exp Aging Res . 2005 ; 31 : 457 - 69 .
9. Patel M , Potter J , Perez I , Kalra L . The process of rehabilitation and discharge planning in stroke: a controlled comparison between stroke units . Stroke . 1998 ; 29 : 2484 - 7 .
10. Shiraishi N , Mizutani C , Menjho M , Deguchi A , Takase K , Hamaguchi H , et al. Comparison of activities of daily living for a convalescent rehabilitation ward and general ward for stroke patients . Nippon Ronen Igakkai Zasshi . 2004 ; 41 : 646 - 52 .
11. Horikawa E , Matsui T , Arai H , Seki T , Iwasaki K , Sasaki H . Risk of falls in Alzheimer's disease: a prospective study . Intern Med . 2005 ; 44 : 717 - 21 .
12. LaBuda J , Lichtenberg P. The role of cognition, depression, and awareness of deficit in predicting geriatric rehabilitation patients' IADL performance . Clin Neuropsychol . 1999 ; 13 : 258 - 67 .
13. Rapport LJ , Webster JS , Flemming KL , Lindberg JW , Godlewski MC , Brees JE , et al. Predictors of falls among right-hemisphere stroke patients in the rehabilitation setting . Arch Phys Med Rehabil . 1993 ; 74 : 621 - 6 .
14. Van Doorn C , Gruber-Baldini AL , Zimmerman S , Hebel JR , Port CL , Baumgarten M , et al. Epidemiology of Dementia in Nursing Homes Research Group. Dementia as a risk factor for falls and fall injuries among nursing home residents . J Am Geriatric Soc . 2003 ; 51 : 1213 - 8 .
15. Mayo NE , Korner-Bitensky N , Levy AR . Risk factors for fracture due to fall . Arch Phys Med Rehabil . 1993 ; 74 : 917 - 21 .
16. Liu-Ambrose T , Khan KM , Eng JJ , Lord SR , McKay HA . Balance confidence improves with resistance or agility training. Increase is not correlated with objective changes in fall risk and physical abilities . Gerontology . 2004 ; 50 : 373 - 82 .
17. Tinetti ME , Williams CS . Falls, injuries due to falls, and the risk of admission to a nursing home . N Engl J Med . 1997 ; 337 : 1279 - 84 .
18. Perell KL , Nelson A , Goldman RL , et al. Fall risk assessment measures: an analytic review . J Gerontol Med Sci . 2001 ;56A: 761 - 6 .
19. Nyberg L , Gustafson Y. Fall prediction index for patients in stroke rehabilitation . Stroke . 1997 ; 28 : 716 - 21 .
20. Vassallo M , Sharma JC , Briggs RS , Allen SC . Characteristics of early fallers in elderly patient rehabilitation wards . Age Ageing . 2003 ; 32 : 338 - 42 .
21. Imai Y , Hsegawa K. The revised Hasegawa's Dementia Scale ( HDS-R ) -Evaluation of its usefulness as a screening test for dementia . J Hong Kong Coll Psychiatr . 1994 ; 4 : 20 - 4 .
22. Kim KW , Lee DY , Jhoo JH , Youn JC , Suh YJ , Jun YH , et al. Diagnostic accuracy of mini-mental status examination and revised hasegawa dementia scale for Alzheimer's disease . Dement Geriatr Cogn Disord . 2005 ; 19 : 324 - 30 .
23. Kwon S , Hartzema AG , Duncan PW , Min-Lai S . Disability measures in stroke: relationship among the Barthel Index, the Functional Independence Measure, and the Modified Rankin Scale . Stroke. 2004 ; 35 : 918 - 23 .
24. Mahoney FI , Barthel D. Functional evaluation: the Barthel Index . Md Med J. 1965 ; 14 : 56 - 61 .
25. Armitage P , Berry G . Statistical methods in medical research, 3rd edn . 2001 ; 355 - 7 . (Japanese translation)
26. Nyberg L , Gustafson Y. Patient falls in stroke rehabilitation. A challenge to rehabilitation strategies . Stroke . 1995 ; 26 : 838 - 42 .
27. Teasell R , McRae M , Foley N , Bhardwaj A . The incidence and consequences of falls in stroke patients during inpatient rehabilitation: factors associated with high risk . Arch Phys Med Rehabil . 2002 ; 83 : 329 - 33 .
28. Luukinen H , Koski K , Laippala P , Kivela SL . Risk factors for recurrent falls in the elderly in long-term institutional care . Public Health . 1995 ; 109 : 57 - 65 .
29. Vassallo M , Sharma JC , Allen SC . Characteristics of single fallers and recurrent fallers among hospital in-patients . Gerontology . 2002 ; 48 : 147 - 50 .
30. Cox DR . Regression models and life tables . J R Stat Soc . 1972 ;B34: 187 - 220 .
31. Kaplan EL , Meier P . Nonparametric estimation from incomplete observations . J Am Stat Assoc . 1958 ; 53 : 457 - 81 .