Content uploaded by Karin Hellström
Author content
All content in this area was uploaded by Karin Hellström on Sep 10, 2014
Content may be subject to copyright.
AUTHOR QUERY FORM
Journal title: CRE
Article Number: 464703
Dear Author/Editor,
Greetings, and thank you for publishing with SAGE. Your article has been copyedited, and we have a few
queries for you. Please respond to these queries when you submit your changes to the Production Editor.
Thank you for your time and effort.
Please assist us by clarifying the following queries:
No Query
1 Please confirm that this is the correct address to be printed for correspondence.
Clinical Rehabilitation
0(0) 1 –7
© The Author(s) 2012
Reprints and permissions:
sagepub.co.uk/journalsPermissions.nav
DOI: 10.1177/0269215512464703
cre.sagepub.com
CLINICAL
REHABILITATION
Background
Falling is a common problem among elderly peo-
ple, especially in patients with stroke.1,2 The risk of
falling is twice as high for patients with stroke than
for people of the same age without stroke.1 Between
5 and 41% of patients admitted to an acute care set-
ting fall at least once,3–5 which makes falling the
most frequent complication during hospitalization
464703CRE0010.1177/0269215512464703Clinical RehabilitationNyström and Hellström
2012
1Vård och bildning, Uppsala kommun, Uppsala, Sweden
2Department of Neuroscience, Section of Physiotherapy,
Uppsala University, Uppsala, Sweden
3Department of Neuroscience, Section of Physiotherapy,
Uppsala University, Uppsala, Sweden
Corresponding author:
Anna Nyström, Uppsala Universitet Hagundagården,
Arkitektvägen 1, 740 20 Vänge, Sweden. [AQ: 1]
Email: karlsson.anna@hotmail.se
Fall risk six weeks from onset
of stroke and the ability of
the Prediction of Falls in
Rehabilitation Settings Tool and
motor function to predict falls
Anna Nyström1,3 and Karin Hellström2
Abstract
Objective: To investigate whether the Prediction of Falls in Rehabilitation Settings Tool (Predict FIRST)
and motor function could be used to identify people at risk of falling during the first six weeks after stroke,
and to compare the risk of falling according to Predict FIRST with real falls frequency.
Design: A longitudinal, prospective study.
Patients: Sixty-eight people newly diagnosed with stroke admitted to an acute stroke unit.
Methods: The participants underwent an assessment of motor ability (Modified Motor Assessment Scale
according to Uppsala University Hospital version 99 (M-MAS UAS-99)) and falls risk (Predict FIRST) on the first
to fourth day at the acute stroke unit. Falls occurring in the acute stroke unit were recorded and falls occurring
after discharge were reported by telephone follow-up. The prediction of falls was analysed with binary logistic regression.
Results: Fourteen of the patients (21%) fell at least once during the first six weeks after stroke. The
strongest significant predictor for falls was a high score on Predict FIRST (odds ratio 5.21, confidence
interval (CI) 1.10–24.78) followed by M-MAS UAS-99 parts C–E (odds ratio 0.65, CI 0.44–0.95). Predict
FIRST underestimated the risk of falling as the median fall risk was 9% according to Predict FIRST.
Conclusion: Although Predict FIRST has the ability to predict falls in people with recent onset of stroke,
there is some underestimation of fall risk.
Keywords
Acute stroke, accidental falls, prevention of falls, prediction of risk, prospective study
Received: 22 May 2012; accepted: 22 September 2012
Article
2 Clinical Rehabilitation 0(0)
after stroke. In community-dwelling stroke survi-
vors, fall incidents are common and the proportion
of fallers are reported to be 34% during the three
months post stroke.6 Falls can lead to a variety of
consequences, such as traumatic brain injuries, frac-
tures, fear of falling, reduced activity and death,7,8
and involve both personal suffering and economic
costs for the community.9
To determine which patients are at increased risk
of falling, fall risk factors in the individuals are usu-
ally assessed.10 Many factors can predict a risk of
falling, and according to Ganz et al.10 these can be
divided into six domains: orthostatic hypotension,
visual impairment, impairment of gait or balance,
medication use, restriction of personal or instru-
mental activities of daily living, and cognitive
impairment. Currently there are no good fall risk
prediction tools available for patients with stroke.
The Prediction of Falls in Rehabilitation Settings
Tool (Predict FIRST) is a simple fall risk index
developed for rehabilitation settings.11 The decisive
factor is whether or not the fall risk index has the
ability to identify people with a risk of falling in an
actual group of patients. The ability of Predict
FIRST to identify people at risk of falling during the
first weeks after stroke has not been investigated to
date,11 nor has motor function measured with the
Modified Motor Assessment Scale according to
Uppsala University Hospital version 99 (M-MAS
UAS-99).
The aims of this study were to investigate
whether Predict FIRST and motor function could be
used to identify people at risk of falling during the
first six weeks after stroke, and to compare the risk
of falling according to Predict FIRST with actual
falls frequency.
Methods
The study had a prospective longitudinal design.
The sample consisted of patients consecutively
recruited after enrolment to an acute stroke unit at
one hospital in the middle of Sweden between 11
February and 1 June 2011. Patients with an acute
stroke are admitted to the acute stroke unit as soon
as the doctors suspect that the patient has a stroke.
Rehabilitation of the patients starts at the acute unit
before they are admitted to post-acute rehabilitation
settings.
Out of 87 eligible patients, 68 with diagnosed
stroke agreed to participate in the study. The exclu-
sion criteria were patients who had difficulty in
understanding or following instructions, patients
who did not speak Swedish, patients diagnosed with
dementia and patients who were not contactable.
The procedures in the study were in accordance
with the ethical standards of the local ethical com-
mittee. The patients were informed about the study
one to four days after admission to the stroke unit.
Written informed consent was obtained from all
participants.
Predict FIRST consists of five fall risk factors,
each giving one point: male, central nervous system
(CNS) medications, a fall in the past year, frequent
toileting, and inability to do tandem stance (i.e.
standing with one foot directly in front of the other
foot). The scale is cumulative (i.e. more risk factors
give a higher risk of falling).11 The probabilities of
falling with different Predict FIRST scores are: 0 =
2%, 1 = 4%, 2 = 9%, 3 = 18%, 4 = 33% and 5 = 52%
risk of falling during the inpatient rehabilitation
period.11
Predict FIRST is a fall risk index with good pre-
dictive validity for use during rehabilitation stay for
patients with different diagnoses who are over 50
years.11 The index is not tested for reliability.
The Modified Motor Assessment Scale accord-
ing to Uppsala University Hospital version 99
(M-MAS UAS-99), Uppsala, Sweden, measures
motor function and mobility.12 For this study, the
M-MAS UAS-99 was divided into the three sub-
scales: M-MAS UAS-99 A–B consisting of back
lying to side lying and from back lying to sitting on
the edge of the bed, C–E covering sitting, sitting to
standing and gait, F–H covering arm functionality,
hand movements and fine motor skills. The reason
for the division was based on the clinical experience
that parts A and B are less demanding for balance
than parts C–E. Parts F–H assess motor skills of the
arm and hand and fine motor skills bilaterally,
which also can influence balance. The highest score
possible for parts A–B was 10, for parts C–E 15 and
parts F–H 30; the maximum score was 55.
Nyström and Hellström 3
The procedure for collection of data was as fol-
lows: the background information (i.e age and use
of walking aid) was filled in by the first author (AN)
and the patient together. The patients were assessed
with the M-MAS UAS-99 by the physiotherapists at
the acute stroke unit on the patient’s second to
fourth day at the acute stroke unit. On the second to
fourth day at the acute stroke unit, all participating
patients were also assessed with Predict FIRST by
the first author in collaboration with the patient’s
doctor and nurses. The doctor answered the ques-
tion about CNS medication, the nurse answered the
question about frequent toileting, the patient or rela-
tive answered the question about falls in the past
year and the author assessed the tandem stance. To
be considered able to do tandem stance, the patient
had to maintain the position without support for 8
seconds.13
In the event of falls occurring in the acute stroke
unit during the study period, a list was established at
the nursing office, this was available to all staff. As
a reminder to reporting any falls, all nurses were
daily on Monday to Friday mornings asked by a
physiotherapist if any falls had occurred. A fall was
defined as unintentionally coming to rest on the
ground or other lower surface without overwhelm-
ing external force or a major internal event (11). We
chose to consider for example a stroke being a
major internal event, whereas, a sudden fall in blood
pressure or dizziness is not.
After discharge from the acute stroke unit, six
weeks after the original date of arrival, phone calls
were made by the first author to the participants to
follow up whether any falls had occurred since dis-
charge. The following questions were asked: ‘Have
you fallen at any time since discharge from hospi-
tal?’ If the patient answered ‘yes’ to that question
the author also asked: ‘How many times have you
fallen since discharge from hospital?’
As data from the M-MAS UAS-99 and its sub-
scales and Predict FIRST were ordinal and they
were not normally distributed, the results from these
instruments are reported with median and range.
Patients who had fallen at least once during the six
weeks after stroke are reported as number and per-
centage. The results of Predict FIRST and real
frequency of falls were reported in number and
percentages. The instrument’s ability to predict falls
was analysed by binary logistic regression. In the
regression model, the results of Predict FIRST were
dichotomized into high or low scores: 0–2 points
was considered a low score and 3–5 points was con-
sidered a high score. In the regression model, the
occurrence of at least one fall was the dependent
variable. M-MAS UAS-99 subscales A–B, C–E and
F–H, Predict FIRST and the number of days the par-
ticipants stayed at the stroke unit were included in
the regression model as independent variables. An
estimation of Nagelkerke was performed to explain
the proportion of the explained variation in the
dependent variable.The level of significance was set
to P ≤ 0.05. Statistical analyses were performed with
SPSS version 20 (SPSS Inc., Chicago, IL, USA).
Results
The sample consisted of 41 men and 27 women
with mean age 72.7 years: participant’s age, length
of stay at the acute stroke unit, side of hemiparesis,
motor function, mobility and walking ability are
presented in Table 1.
Of the 68 participants, 14 patients (21%: 5
women and 9 men) fell at least once during the
study period. Of the 5 women who fell during the
study period, 4 fell once and 1 fell six times and of
the 9 men who fell during the study period, 2 fell
once, 3 fell twice and 4 fell three times or more. In
total, 36 falls occurred, women fell 10 (28%) times
and men fell 26 (72%) times. At the acute stroke
unit, there were 23 falls and after discharge there
were 13 falls.
The participants’ fall risk according to Predict
FIRST demonstrated a median of 2 points (i.e. a 9%
probability of falling). For Predict FIRST 21 (31%)
participants scored 2 points, none scored 5 points,
and 3 participants scored 0 points. The distribution
of fall risks according to Predict FIRST is presented
in Table 2. Of the 7 participants estimated to have a
33% risk of falling in Predict FIRST, 6 fell (Table 2).
This meant a real fall frequency of 85.7% in this
group. Fall risks of 4%, 9% and 18% according to
Predict FIRST and the real frequency of falls
corresponded as the real fall frequency was only
4 Clinical Rehabilitation 0(0)
slightly higher than that estimated by Predict FIRST.
Only 3 participants had an estimated fall risk of 2%
but none of them fell.
As seen in Table 3 the strongest significant
variable for predicting falls was the sum of
Predict FIRST with odds ratio 5.21 (CI 1.10–
24.78). The other significant variable was the
mobility part, C–E, in the Modified Motor
Assessment Scale with odds ratio 0.65 (CI 0.44–
0.95). The other variables were non-significant.
The regression model explained 43% of the vari-
ance of at least one fall during the first six weeks
after stroke.
Discussion
The main, novel finding of the present study was the
demonstrated ability of the fall prediction tool
Predict FIRST, with odds ratio of 5.21, to predict
falls in the first six weeks from onset of stroke.
Mobility was also a significant predictive factor for
a fall, with odds ratio of 0.65. The regression model
explained 43% of the variance for at least one fall
during the first six weeks after stroke.
One reason why Predict FIRST was shown to be
the variable with the strongest predictive ability
may be because the instrument was developed
Table 2. Distribution of subjects in the study group according to risk of falling, assessed by Predict FIRST, and the
number of patients fallen/not fallen within each risk group (n = 68)
Predict First: Points/
Estimated fall risk (%)
0 (2%) 1 (4%) 2 (9%) 3 (18%) 4 (33%) 5 (52%) Total
Number with estimated
fall risk according to
Predict First points (%)
3 (4.4) 19 (27.9) 21 (30.9) 18 (26.5) 7 (10.3) 0 (0) 68 (100)
Number who have fallen (%) 0 (0) 1 (5.3) 3 (14.3) 4 (22.2) 6 (85.7) 0 (0) 14 (20.6)
Number who have not
fallen (%)
3 (100) 18 (94.7) 18 (85.7) 14 (77.8) 1 (14.3) 0 (0) 54 (79.4)
Table 1. Mean age, length of stay at the stroke unit, side of hemiparesis, results on M-MAS UAS-99 and walking
ability of the sample (n = 68)
Age M (SD) Min–Max 72.7 (11.41) 43–98
Number of days at the stroke unit M (SD) Min–Max 10.7 (8.43) 3–35
Hemiparesis on the right side 18 (26)
Hemiparesis on the left side 19 (28)
No difference between the right and left side in
motor performance
31 (46)
M-MAS UAS-99
Back lying to side lying and back lying to sitting on
the edge of the bed (max 10 points)
Median 10 (Min–Max 2–10)
Sitting, sitting to standing and gait (max 15 points) Median 13 (Min–Max 1–15)
Arm functionality, hand movements and fine
motor skills (max 30 points)
Median 28 (Min–Max 5–30)
Total (max 55 points) Median 50.5 (Min–Max 8–55)
Walking without aids, number (%) 32 (46.4)
Walking with aids, number (%) 26 (37.7)
Cannot walk, number (%) 10 (14.5)
M-MAS UAS-99, Modified Motor Assessment Scale according to Uppsala University Hospital version 99.
Nyström and Hellström 5
specifically to assess the risk of falling, and it
addresses several falls risk factors, in contrast to the
Modified Motor Assessment Scale, which only
measures motor skills. Parts C–E, sitting, sitting to
standing and gait, is the subscale that is the most
demanding for balance and may be the reason why
this particular subscale of the M-MAS UAS-99 was
a significant variable in the regression model. This
subscale had an odds ratio below 1 (odds ratio
0.645), which meant low scores in sitting, sitting to
standing and gait, was a protective factor against
falls. For example, if a person is less independent in
these activities, they receive more help and support
from the ward staff, which is protective against
falls. The incidence of fall was reported to be high-
est in patients with severe walking impairments and
who received early walking training, although clini-
cally relevant improvements in walking ability were
achieved.14 The process of gaining mobility post
stroke appears associated with a higher risk for falls.
In a study by Persson et al.,15 the Modified Motor
Assessment Scale according to Uppsala University
Hospital version 95 was able to identify patients at
risk of falling during the first year after stroke, but
in that study, a score under 50 meant a risk of
falling.15
The variables in the regression model explained
43% of the variance of at least one fall the first six
weeks after stroke. This meant the remaining 57%
was explained by other factors, which highlighted
falling is a multidimensional problem, and the
value of a predictive model can be increased with
variables other than just purely physical variables.2
In a literature review on fall risk factors in stroke
patients16 gender does not matter. However, in the
study when Predict FIRST was developed11 and in
a study by Nyberg and Gustafson,17 men have a
greater risk of falling. In our study, the difference
between men and women falling was only mar-
ginal, as approximately 18.5% of women and 22%
of the men fell at least once. The number of falls
was more than doubled in men as seven men fell
more than once, in contrast to just one woman who
fell repeatedly.
The majority (85.3%) of patients had between 1
and 3 points on Predict FIRST, corresponding to a
risk of falling between 4% and 18%. The risk of
falling according to Predict FIRST was close to the
real fall frequency up to an estimated risk of 18%.
When the calculated risk of falling was 33%,
according to Predict FIRST, the real fall frequency
was much higher (85.7%). The real fall frequency
was also higher than the estimated probability of
4%, 9% and 18% risk of falling. Only patients with
a risk of 2% demonstrated a lower real fall fre-
quency. Thus, in patients newly diagnosed with
stroke, Predict FIRST underestimated the patients’
risk of falling in the current study. As no patient in
this study had a fall risk of 52% according to Predict
FIRST, it was not possible to compare this estimated
fall risk with the real fall frequency. One reason the
fall risk according to Predict FIRST did not fully
comply with the real fall frequency may be that
patients with stroke have a greater risk of falls than
other groups of patient present at the rehabilitation
department where Predict FIRST was tested.11
Table 3. Binary logistic regression analysis of fall risk factors, with the odds ratios, confidence intervals and
significance level for the independent variables presented
Independent variables Risk of falling Odds ratio (95%
confidence interval)
P-value
Number of days at the stroke unit 1.09 (0.99–1.20) 0.093
M-MAS UAS-99 A–B 1.78 (0.94–3.39) 0.078
M-MAS UAS-99 C–E 0.65 (0.44–0.95) 0.026
M-MAS UAS-99 F–H 1.11 (0.90–1.37) 0.341
Predict FIRST risk of falling ≤9% = 0; ≥18% = 1 5.21 (1.10–24.78) 0.038
Nagelkerke R2=0.43.
M-MAS UAS-99, Modified Motor Assessment Scale according to Uppsala University Hospital version 99.
6 Clinical Rehabilitation 0(0)
The participants in the present study had a median
fall risk of 9% according to Predict FIRST, which was
low compared to reported frequencies of falls in other
studies,3–5,17–19 where the frequency of falls is between
5 and 41% in the first months after a stroke. The study
reporting the lowest percentage of patients falling was
in an acute setting (5%) and should be similar to the
setting our study was conducted in. In the present
study, 20.6% of patients fell at least once, which was
four times higher than the reported 5% in the study by
Schmid et al.5 There is some evidence that people with
stroke are more likely to fall in the early stages of reha-
bilitation, particularly in the first week.20,21
There are some limitations in the present study. Falls
occurring during the patients’ stay at the acute stroke unit
were reported by the staff either orally or in writing.
Although there were several ways to report falls to the
study and it was made as easy as possible to report falls,
there might still have been a small number of unreported
falls. Also, an eventual recall bias at the phone follow-up
might have reduced the number of falls reported.
Further studies investigating the relationship
between the risk of falling according to Predict First
and the real risk of falling in patients in different
stages after stroke are required to verify the increased
real fall risk identified in this study. In addition,
Predict FIRST should be tested for reliability.
In conclusion, Predict FIRST can be used in the
assessment of fall risk in patients newly diagnosed
with stroke: however, the real frequency of falls was
higher than the risk of falling estimated by Predict
FIRST. Despite this, a fall risk index that is easy to
use should be routine for a stroke unit for directing
staff attention to fall prevention in a group of
patients with a high risk of falling.
Funding
This research received no specific grant from any funding
agency in the public, commercial or not-for-profit
sectors.
Conflict of interest
The authors declare that there is no conflict of interest.
References
1. JØrgensen L, Engstad T and Jacobsen B. Higher incidence
of falls in long-term stroke survivors than in population
controls: depressive symtoms predict falls after stroke.
Stroke 2009; 33: 542–547.
2. Baetens T, De Kegel A, Calders P, et al. Prediction of fall-
ing among stroke patients in rehabilitation. J Rehabil Med
2011; 43: 876–883.
3. Holloway RG, Tuttle D, Baird T, et al. The safety of hospi-
tal stroke care. Neurology 2007; 68: 550–555.
4. Tutuarima JA, Van der Meulen JH, De Haan RJ, et al. Risk
factors for falls of hospitalized stroke patients. Stroke 1997;
28: 297–301.
5. Schmid AA, Wells CK, Concato J, et al. Prevalence, pre-
dictors, and outcomes of poststroke falls in acute hospital
setting. J Rehabil Res Devel 2010; 47: 553–562.
6. Smith J, Forster A and Young J. Use of the ‘STRATIFY’
falls risk assessment in patients recovering from acute
stroke. Age Ageing 2006; 35: 138–143.
7. Friedman SM, Munoz B, West SK, et al. Falls and fear of
falling: which comes first? A longitudinal prediction model
suggest strategies for primary and secondary prevention.
JAGS 2002; 50: 1329–1335.
8. Holmgren E. Getting up when falling down. Reducing fall
risk factors after stroke through an exercise program.Dis-
sertation, Umeå Universitet, Sweden, 2010.
9. Batchelor FA, Mackintosh SF, Said CM and Hill KD. Falls
after stroke. Int J Stroke 2012; 7: 482–490.
10. Ganz DA, Bao Y, Shekelle PG, et al. Will my patient fall?
JAMA 2007; 297: 77–86.
11. Sherrington C, Lord SR, Close JCT, et al. Development of a tool
for prediction of falls in rehabilitation settings (Predict FIRST):
a prospective cohort study. J Rehabil Med 2010; 42: 482–488.
12. Barkelius K, Johansson A, Kaoken K, et al. [Reliability and
validity testing of the modified Motor Assessment Scale
according to Uppsala University Hospital-95]. Nord Fys-
ioter 1997; 1: 121–126 (in Swedish).
13. Jonsson E, Seiger Å and Hirschfeldt H. Postural steadi-
ness and weight distribution during tandem stance in
healthy young and elderly adults. Clin Biomech 2005;
20: 202–208.
14. Tilson JK, Wu SS, Cen SY, et al. Characterizing and iden-
tifying risk for falls in the LEAPS study: a randomized
clinical trial of interventions to improve walking poststroke.
Stroke 2012; 43: 446–452.
Clinical messages
•We suggest the use of Predict FIRST as a sig-
nificant predictor for falls the first six weeks
after stroke.
•Twenty-one per cent of the patients fell at least
once during the first six weeks after stroke.
•The frequency in number of falls was more
than doubled in men with stroke in the first
six weeks after stroke.
Nyström and Hellström 7
15. Persson C.U, Hansson P-O and Sunnerhagen K.S. Clinical
tests performed in acute stroke identify the risk of falling
during the first year: postural stroke study in Gothenburg
(POSTGOT). J Rehabil Med 2011; 43: 348–353.
16. Campbell GB and Matthews JT. An integrative review of
factors associated with falls during post-stroke rehabilita-
tion. J Nurs Scholarsh 2010; 43: 395–404.
17. Nyberg L and Gustafson Y. Fall prediction index for
patients in stroke rehabilitation. Stroke 1997; 28: 716–721.
18. Nyberg L and Gustafsson Y. Using the Downton index to
predict those prone to falls in stroke rehabilitation. Stroke
1996; 27: 1821–1824.
19. Nakagawa Y, Sannomiya K, Kinoshita M, et al.
Development of an assessment sheet for fall predic-
tion in stroke inpatients in convalescent rehabilitation
wards in Japan. Environ Health Prevent Med 2008;
13: 138–147.
20. Suzuki T, Sonada S, Misawa K, Saitoh E, Shimizu Y and
Kotake T. Incidence and consequence of falls in inpatient
rehabilitation of stroke patients. Exp Aging Res 2005; 31:
457–469.
21. Rabadi MH, Rabadi FM and Peterson M. An analysis of
falls occuring in patients with stroke on an acute rehabilita-
tion unit. Rehabil Nurs 2008; 33: 104–109.