Received: 8 December 2019
|
Revised: 12 May 2020
|
Accepted: 17 May 2020
DOI: 10.1111/jocn.15387
ORIGINAL ARTICLE
Comparison of the predictive validity of three fall risk
assessment tools and analysis of fall-risk factors at a tertiary
teaching hospital
Eun Hee Cho MSN, RN, CNS, Nurse1
| Yun Jung Woo MSN, RN, Nurse2
|
3
4
Arum Han MSN, RN, Nurse
| Yoon Chung Chung MSN, RN, Nurse
|
5
Yeon Hee Kim PhD, RN, Associate Professor
| Hyeoun-Ae Park PhD, RN, Professor6
1
Emergency Department, Department of
Nursing, Asan Medical Center, Seoul, Korea
Abstract
2
Aims and objectives: The main purpose of this study was to identify the best fall-risk
Oncology Team, Department of Nursing,
Asan Medical Center, Seoul, Korea
3
Outpatient Nursing Team, Department of
Nursing, Asan Medical Center, Seoul, Korea
4
Department of Nursing, Asan Medical
Center, Seoul, Korea
5
Department of Clinical Nursing, Graduate
School of industry, University of Ulsan,
Seoul, Korea
6
College of Nursing and Research Institute
of Nursing Science, Seoul National
University, Seoul, Korea
Correspondence
Hyeoun-Ae Park, College of Nursing and
Research Institute of Nursing Science, Seoul
National University, 103 Daekak-ro, Jongrogu, Seoul 03080, Korea.
Email: hapark@snu.ac.kr
Funding information
This work was supported by the Nursing
Department of Seoul Asan Medical Center
as a policy research.
assessment tool, among the Morse Fall Scale, the Johns Hopkins fall-risk Assessment
Tool and the Hendrich II fall-risk Model, for a tertiary teaching hospital. The study
also analysed fall-risk factors in the hospital, focusing on the items of each fall assessment tool.
Methods: Data on falls were obtained from the patient safety reports and electronic
nursing records of a tertiary teaching hospital. A retrospective study was conducted
to compare the sensitivity, specificity, area under the curve, positive predictive value,
negative predictive value, Youden index and accuracy of the Morse Fall Scale, the
Johns Hopkins fall-risk Assessment Tool and the Hendrich II fall-risk Model. This
study was conducted according to the Strengthening the Reporting of Observational
Studies in Epidemiology guideline for reporting case–control studies.
Results: By analysing the association between falls and the items included in the
three tools, we identified significant fall-risk factors such as gait, dizziness or vertigo,
changes in mental status, impulsivity, history of falling, elimination disorder, drugs
affecting falls, and depression.
Conclusions: The Hendrich II fall-risk Model had the best predictive performance
for falls of the three tools, considering the highest in the area under the curve and
the Youden index that comprehensively analysed sensitivity and specificity, while
the Johns Hopkins fall-risk Assessment Tool had the highest accuracy. The most significant fall-risk predictors are gait, dizziness or vertigo, change in mental state, and
history of falling.
Relevance to clinical practice: To improve the fall assessment performance of the
Morse Fall Scale at the study hospital, we propose that it be supplemented with four
most significant fall-risk predictors identified in this study.
3482
|
© 2020 John Wiley & Sons Ltd
wileyonlinelibrary.com/journal/jocn
J Clin Nurs. 2020;29:3482–3493.
|
CHO et al.
3483
KEYWORDS
falls, patient safety, risk assessment, risk factors
1 | I NTRO D U C TI O N A N D BAC KG RO U N D
Patient falls are unexpected events in healthcare institutions that
can lead to physical injury or even death, as well as negatively
impact the treatment process. Falls may also result in a prolonged
hospital stay, incur additional medical costs, or even lead to medical malpractice lawsuits (Dunne, Gaboury, & Ashe, 2014; Hong,
Kim, Jin, Piao, & Lee, 2015; Jung & Park, 2018). Inpatients admitted to acute care hospitals may have many different health
issues and potential fall-risk factors (Jang & Lee, 2014; Jung &
Park, 2018). In light of this, it is necessary to take measures to
prevent falls in acute care settings in order to support patients in
What does this paper contribute to the wider
global clinical community
• This study was conducted in an acute setting hospital in
which patients with various characteristics are admitted.
• In addition, this study identified significant fall-risk factors, regardless of fall-risk assessment tools.
• Regardless of the fall-risk assessment tool used in acute
setting hospital, we recommend to consider gait, dizziness
or vertigo, mental status or impulsivity, and fall history
which are identified as significant risk factors in this study.
their recovery and enable medical staff to concentrate on treatment. Patient falls can be prevented by assessing their fall-risk.
Various fall-risk assessment tools have been developed to identify
patients who are at high fall-risk (Kim & Choi-Kwon, 2013). There
have been various studies to develop these tools and test their
2 | M E TH O DS
2.1 | Purpose
validity. The fall-risk tools used in acute care hospitals include the
Morse Fall Scale (MFS), the Johns Hopkins fall-risk Assessment
The purpose of this study was to compare the sensitivity, specificity,
Tool (JHFRAT) and the Hendrich II fall-risk Model (HFRM)
area under the curve (AUC), positive predictive value, negative pre-
(Caldevilla, Costa, Teles, & Ferreira, 2013; Poe et al., 2018). The
dictive value, Youden index and accuracy of the MFS, JHFRAT and
predictive accuracy of these assessment tools may vary by insti-
HFRM to determine which among them has the highest predictive
tution due to different patient characteristics, socio-economic
validity for fall-risk in the study hospital and to examine whether the
factors and physical environments (Aranda-Gallardo et al., 2013;
items in each tool were related to fall occurrences.
Baek, Piao, Jin, & Lee, 2014; Hou et al., 2017); thus, it is important to identify and use the most appropriate fall-risk assessment tool for each institution (Aranda-Gallardo et al., 2013; Baek
2.2 | Setting
et al., 2014; Hou et al., 2017; Vassallo, Stockdale, Sharma, Briggs,
& Allen, 2005).
There is a wide range of fall-risk factors, such as musculoskel-
The study hospital is a tertiary teaching hospital with 2,600 beds for
patients requiring acute care. The hospital occupancy rate is over 95%.
etal weakness, balance impairment, fall history, urination/elimination disorder, medication that increase fall-risk, central nervous
system depressants and stimulants (e.g., benzodiazepines, narcot-
2.3 | Research design
ics, antiepileptic drugs, cardiovascular drugs), and chronic diseases
including cancer (Callis, 2016; Hou et al., 2017). According to a
This is a retrospective study using fall occurrence data docu-
study conducted with the MFS in South Korea, some factors in the
mented in the PSRs of the study hospital and the items of the MRS,
assessment tool had no significant relevance to patient falls, which
JHFRAT and HFRM from the ENRs of the study hospital. This study
undermines the tool's validity (Jang & Lee, 2014). It is important,
was conducted according to the Strengthening the Reporting of
therefore, for each healthcare institution to identify whether the
Observational Studies in Epidemiology (STROBE) guideline for re-
items in the fall-risk assessment tool are in line with the actual fall-
porting case–control studies (Appendix S1).
risk factors. The MFS is being used to assess fall-risk in the study
hospital; thus, its predictive validity for this hospital needed verification in comparison with other fall-risk assessment tools. We
compared the predictive performance of the MFS against that of
the JHFRAT and the HFRM by utilising the patient safety reports
2.4 | Study subjects
2.4.1 | Case group
(PSRs) and electronic nursing records (ENRs). We also analysed
whether the items of each tool had significant relevance to the pa-
The case group included 447 adult inpatients aged 19 or over who had
tient falls.
fallen according to the PSR submitted to the Quality Improvement
3484
|
Department of the study hospital between 1 June 2014–31 May
CHO et al.
2.5.3 | Hendrich II fall-risk model
2015. The paediatrics, psychiatric, intensive care unit, emergency
and outpatient departments were excluded because they do not use
The HFRM was developed to analyse the fall-risk factors of acute
the MFS as a fall assessment tool.
care hospitals by Hendrich in 1988 and revised in 1995 (Han
et al., 2017; Hendrich, Bender, & Nyhuis, 2003) (Appendix 3). The
2.4.2 | Control group
tool consists of eight items: confusion/disorientation/impulsivity,
depression, altered elimination, dizziness or vertigo, gender (male),
antiepileptics, benzodiazepines and the Get-Up-and-Go test. The
The control group included 1,341 patients, three times the number
total score can range from 0–16. The score for being considered at
of subjects in the case group, who were admitted to the study hos-
high-risk in this study was five or greater.
pital during the same period but did not fall. The case:control ratio
of 1:3 was chosen based on Grimes and Schulz (2005) and G-power
statistics with effect size of 0.2 and power of 0.9. They were se-
HFRM items were retrospectively extracted from the ENRs.
The Get-Up-and-Go test in this tool is divided into 0, 1, 3 and 4,
and each criterion was defined in this study as follows:
lected based on their age, gender, surgery, department and fall-risk
assessment score at the time of admission matched to those of the
case group.
• 0 represents the following characteristics: alert, not dizzy, not
weak in the lower limbs, mobile, and has no impaired gait.
• 1 represents the following characteristics: alert but dizzy, or has a
2.5 | Measuring instruments
2.5.1 | Morse fall scale
“weak” gait in the MFS but does not use an ambulatory aid.
• 3 represents the following characteristics: alert, a “weak” or “impaired” gait in the MFS, and uses either “crutches/cane/walker” or
“furniture or others” as an ambulatory aid.
• 4 represents the following characteristics: not alert, a “weak”
The MFS was developed and validated in acute care, long-term care
or “impaired” gait in the MFS, and uses either “crutches/cane/
and rehabilitation by Morse in 1989 (Han et al., 2017; Morse, Black,
walker” or “furniture or others” as an ambulatory aid.
Oberle, & Donahue, 1989) (Appendix 1). The tool consists of six
items: history of falling, secondary disease, ambulatory aid, intravenous therapy/heparin lock, gait and mental status. The total score
2.6 | Ethical considerations
can range from 0–125. The score for being considered at high-risk in
the study hospital was 45 or greater.
This study was approved by the Institutional Review Board of the
study hospital. To make the data unidentifiable, the registration
2.5.2 | The johns hopkins fall-risk assessment tool
numbers were replaced by random numbers. The data file was encrypted, and the data were accessed only by the two co-authors.
The data will be deleted 3 years after the end of the study, when the
The JHFRAT was developed by the Johns Hopkins Hospital
mandatory record retention period expires.
in 2005 and revised in 2007 (Poe, Cvach, Gartrelu, Radzik, &
Joy, 2005; Poe et al., 2018) (Appendix 2). The tool consists of 7
items: age, fall history, elimination, medication, patient care de-
2.7 | Data collection
vices, mobility and cognition. The total score can range from 0–34.
The score for being considered at high-risk in this study was 14 or
The data used in this study were collected and reviewed by the two
greater.
co-authors. To maximise inter-rater reliability, the authors rated the
JHFRAT items were retrospectively extracted from the ENRs.
items of the three fall-risk assessment tools and reviewed the scores
The following criteria were utilised to compare the items of the
on 30 samples until weighted kappa reached 0.8 prior to collecting
JHFRAT with those of the MFS:
the data.
• Past fall history in the JHFRAT was limited to fall occurrences
from the PSRs from 1 June 2014–31 May 2015. The age, gender,
Occurrence and date of fall of the case group were extracted
within the past 3 months, as defined in the MFS.
surgery experience, department, MFS fall-risk score at the time of
• The “requires assistance or supervision for mobility, transfer,
admission, and the items of the three fall-risk assessment tools were
or ambulation” mobility item in the JHFRAT was defined as the
extracted from nursing assessment records, fall-risk assessment re-
“crutches/cane/walker” or “furniture/other aid” ambulatory aid
cords, medication administration records, nursing progress notes,
item in the MFS.
nursing assessment notes and the order communication system. The
• The “unsteady gait” mobility item in the JHFRAT was defined as
the “impaired” or “weak” gait item in the MFS.
items of the three fall-risk assessment tools in the case group were
collected from the ENR documented on the date of fall (Table 1).
|
CHO et al.
TA B L E 1 Data sources for the items of
three tools
Fall-risk assessment tools
data sources
3485
MFS
JHFRAT
HFRM
Nursing information
records
Secondary disease
Age
Elimination
Sex
Altered
elimination
Nursing records
Mental status
Cognition
Confusion/
disorientation/
impulsivity
Dizziness
Symptomatic
depression
Order Communication
System (OCS)
Nursing treatment records
Intravenous
therapy/Heparin
lock
Patient care
equipment
Medication records
Drugs affecting
falls
Laboratory nursing
records
Sedated procedure
within past 24 hr
Fall-risk assessment
records
Antiepileptics
Benzodiazepines
History of falling
Gait
Ambulatory aid
Fall history
Mobility
Get-up-&-go test
high-risk score ≥ 45
high-risk score > 13
high-risk score ≥ 5
The items of the three tools in the control group were collected
Improvement Department during the study period. The mean age of
from the ENR documented on the same date, if the matched subject
the case group was 61 years, 59.8% of them were 60 years or over,
in the case group fell within 10 days of admission to the hospital.
and 54.6% of them were male. The mean number of days between
The items of the three tools in the control group were collected
the fall occurrence and admission was 21.94, with 52.35% of them
from the ENR documented on the 10th day after admission, if the
having fallen within 10 days after admission. Approximately 25% of
matched subject in the case group fell after 10 or more days since
them were from hemato-oncology, followed by 20% from gastroen-
admission to the hospital. This was due to the shorter length of hos-
terology and 20% from general surgery; approximately 62.6% had
pital stay in the control group.
not had any surgery. Based on the fall-risk score at the time of admission, 71% of them were in the low fall-risk group.
2.8 | Analysis
There were 1,341 individuals in the control group, three times the
number of subjects in the case group, matched on the demographic
characteristics of the control group. The lack of statistical difference
The data were analysed with SPSS/WIN 20.0 and SAS 9.4. The homo-
in the general characteristics between the two groups provides evi-
geneity of the age, sex, surgery experience, department and MFS score
dence that the groups were adequately matched (Table 2).
at admission was tested with the chi-squared test using a significance
cut-off of 0.05. The predictive performances of three fall-risk assessment tools were compared using the odds ratio, sensitivity, specificity,
positive predictive value, negative predictive value, Youden index and
3.2 | Predictive performance comparison of the fallrisk assessment tools
AUC. Univariate analyses and multivariate analyses were performed
to examine the relationship between the items of the three fall-risk
There were more falls in the highrisk groups than the lowrisk
assessment tools and fall occurrences in the study hospital. Finally,
groups predicted by all the three fall-risk assessment tools. Most
how the items in each fall-risk assessment tool were related to fall oc-
falls occurred in the highrisk group predicted by the JHFRAT
currence was analysed using conditional logistic regressions.
(48.5%), followed by HFRM (46.5%) and the MFS (36.1%). The
least number of falls occurred in the lowrisk group predicted by
3 | R E S U LT S
3.1 | General characteristics
the HFRM (15.4%), followed by MFS (17.3%) and the JHFRAT
(20.9%). The odds ratio of the HFRM's highrisk group was 7.41
times higher than those of the other two tools. Regarding the predictive validity of the three tools, the MFS was highest (59.28%)
in sensitivity (the ratio of patients assessed by the tool as being
There were 447 individuals in the case group aged 19 and older
highrisk among those who fell), whereas the JHFRAT was highest
who had fallen according to the PSR submitted to the Quality
(89.71%) in specificity (the ratio of patients categorised by the tool
|
3486
Characteristics
CHO et al.
Faller (%)
(n = 447)
Non-faller (%)
(n = 1,341)
TA B L E 2 General characteristics and
the Morse Fall Scale of the study subjects
2
P
1.0 (0.99, 1.00)
0.04
1.00
0.83 (0.82, 0.83)
0.06
.83
0.50 (0.49, 0.51)
<0.001
1.00
0.22 (0.21, 0.23)
1.52
.23
0.11 (0.10, 0.11)
2.84
.11
OR (95% CI)
χ
Age (years)
<60
180 (40.2)
537 (39.9)
60 ~ 69
134 (30.0)
399 (29.8)
70 ~ 79
100 (22.4)
303 (22.6)
≤80
33 (7.4)
102 (7.6)
Gender
Female
203 (45.4)
600 (44.7)
Male
244 (54.6)
741 (55.3)
Fall occurrence after admission
within 10 days
234 (52.3)
After 10 days
213 (47.7)
Discipline
HEM, ONC
112 (25.1)
336 (25.1)
GI
89 (19.9)
267 (19.9)
GS
87 (19.5)
261 (19.5)
NS, NR, RM
35 (7.8)
105 (7.8)
OBY, URO
31 (6.9)
93 (6.9)
a
30 (6.7)
90 (6.9)
CV
22 (4.9)
66 (4.9)
IM
CS, OS, PS
19 (4.3)
57 (4.3)
PLM
18 (4.0)
54 (4.0)
4 (0.9)
12 (0.9)
ENT, OPH
Operation during hospital stay
Yes
167 (37.4)
458 (34.2)
No
280 (62.6)
883 (65.8)
Risk of Morse fall scale at admission
a
Low-risk
318 (71.1)
1,008 (75.2)
High-risk
129 (28.9)
333 (24.8)
ALG, DER, END, FM, GIM, INF, NPH, RHE
as lowrisk among those who did not fall). In terms of positive pre-
falls; mental status (OR = 5.90) was the most likely to affect patient
dictive value (the ratio of those who fell among those categorised
falls, followed by gait (OR = 3.30) and ambulatory aid (OR = 3.09).
by the tool as highrisk), the JHFRAT was best (48.51%). In terms
Furthermore, these same items were found to have a significant im-
of negative predictive value (the ratio of those who did not fall
pact on patient falls in the multivariate analyses, with the risk especially
among those categorised by the tool as lowrisk), the HFRM was
correlated with mental status (OR = 4.61) or gait (OR = 2.45) issues.
best (84.56%). The AUC of the HFRM was 74.2%, higher than that
In the univariate analyses, fall history, elimination, number of
of the MFS (64.1%) and the JHFRAT (70.8%). The Youden index
fall-related drugs, three items related to mobility, and three items
of the HFRM was also higher (0.35) than that of the MFS (0.24)
related to cognition in the JHFRAT were significant predictors of
and the JHFRAT (0.19). The accuracy of the JHFRAT was highest
patient falls. The “requires assistance or supervision” mobility item
(74.55%), followed by the HFRM (72.82%) and the MFS (63.54%)
(OR = 14.74) was the most likely to affect patient falls, followed by
(Figure 1, Table 3).
the “impulsive” cognition item (OR = 9.95) and the “visual or auditory
impairment” mobility item (OR = 5.87). The multivariate analyses
3.3 | Relationship between assessment
items and falls
showed that fall history, elimination, three items related to mobility,
and two items related to cognition in the JHFRAT had a significant
impact on patient falls, with the “requires assistance or supervision”
mobility item being the most important (OR = 10.39), followed by
In the univariate analyses, the fall history, ambulatory aid, gait and
the “visual or auditory impairment” mobility item (OR = 5.52) and the
mental status items on the MFS were significant predictors of patient
“impulsive” cognition item (OR = 4.46).
|
CHO et al.
3487
Matarese, and Pedone (2011), which examined the HFRM's predictive validity on elderly inpatients in an acute care setting, the AUC
was 72%, also similar to the findings of this study.
In the current study, the HFRM's sensitivity was similar to that
of the MFS, which is used in the study hospital, with a difference of
2%; however, the accuracy and specificity of the HFRM were higher
than those of the MFS. It is notable that the fall ratio in the low-risk
group was 15.4% for the HFRM, lower than that of the MFS (17.3%)
and JHFRAT (20.8%), showing that the validity of the HFRM is the
best in the low-risk group. In light of this, it is necessary to analyse
the factors that played a role in the low-risk group's fall, and further
studies should focus on fall prevention in the low-risk group.
Univariate analyses were conducted to examine the relationship
between the items of the three assessment tools and actual patient
falls, finding that gait, dizziness or vertigo, changes in mental status
and impulsivity, fall history, elimination disorder, medications and
depression were significant risk factors. Multivariate analyses were
conducted to examine the relationship between the items of the
three assessment tools and patient falls. It was found that gait, dizziFIGURE 1
Comparison of ROC curves of three tools
ness or vertigo, changes in mental status and impulsivity, fall history,
elimination disorder, medications and depression were identified as
fall-risk factors. Among the significant factors in univariate analyses,
Additionally, all seven items of the HFRM were significantly
the four items with the highest ORs (gait, dizziness or vertigo, mental
related to patient falls in the univariate analyses, with dizziness
status and impulsivity, and fall history) will be discussed compared
or vertigo (OR = 12.04) being the most important, followed by al-
with the result of the multivariate analyses.
tered elimination (OR = 3.88) and poor performance on the Get-
First, gait is a significant fall-risk factor item included in all three
Up-and-Go test (OR = 3.34). In the multivariate analyses, four items
tools. In particular, fall-risk was 14.7 and 4.3 times higher for those
(symptomatic depression, altered elimination, dizziness or vertigo,
who “require assistance or supervision” and have an “unsteady gait”
and Get-Up-and-Go test performance) were significantly related to
in the JHFRAT, respectively. In the other two tools, patients with gait
patient falls, with dizziness or vertigo (OR = 4.83) being the most
problems were more than three times more likely to fall. This is con-
important, followed by altered elimination (OR = 2.19) and poor per-
sistent with the crucial nature of gait as a risk factor for falls found in
formance on the Get-Up-and-Go test (OR = 2.03).
previous studies (e.g., Eagles, Yadav, Perry, Sirois, and Emond (2018);
The multivariate analyses of the goodness-of-fit of the three
Kim, Lee, and Eom (2008); Yang and Chun (2009)). In clinical setting,
tools indicated that the JHFRAT offered the best fit (AIC = 754.82;
gait is assessed indirectly by interviewing a patient or caregiver more
Table 4).
often than by direct observation by a nurse. Based on a systematic
review, Eagles, Yadav, Perry, Sirois, and Emond (2018) found that gait
4 | D I S CU S S I O N
self-assessment is unreliable. In addition, Eagles et al. (2018) found
that Time Up and Go (TUG) test is a widely used mobility measure
and takes less than a minute to perform in general. Therefore, it is
In this study, the predictive performance of the MFS, which is used
necessary for the nurse to directly assess the patient's gait using an
in the study hospital, was compared with that of the JHFRAT and
accessible and effective gait evaluation tool such as TUG. To pre-
HFRM, and the relationship between the items of the three fall-risk
vent patient falls, it is also necessary to identify physical problems
assessment tools and fall occurrences was analysed. The predictive
that affect patient's gait, such as visual or auditory impairments, and
performance of the three tools was compared based on a cut-off
adopt appropriate interventions, such as training in the correct use
score for being considered at high-risk of falling. While the differ-
of ambulatory aids or mobility assistance/supervision.
ence was not considerably substantial, the HFRM showed higher
Second, dizziness or vertigo is a risk factor exclusive to the HFRM.
predictive performance in the Youden index and AUC regarding sen-
fall-risk was found to increase 12 times when patients had dizziness
sitivity and specificity than the other two tools. This result is consist-
or vertigo. Harrison, Ferrari, Campbell, Maddens, and Whall (2010);
ent with that of study by Kim, Mordiffi, Bee, Devi, and Evans (2007),
and Kim, Lee, and Eom (2008) found that the presence of dizziness
which compared the fall-risk assessment tools used on inpatients
or vertigo led to a 5.1 and 7.1 times increased fall-risk, respectively.
in an acute care setting; the HFRM's predictive performance was
Ambrose, Paul, and Hausdorff (2013) found that dizziness or vertigo
higher than that of the MFS and STRATIFY, with an AUC of 73%,
could lead to patient falls by making posture or gait unsteady. It is
similar to the findings of this study. Moreover, in a study by Ivziku,
therefore important to note that any change in posture or gait may
|
CHO et al.
lead to falls in patients experiencing dizziness or vertigo. The antiepileptics HFRM item was a fall-risk factor in the univariate analyses
72.82
74.55
63.54
Accuracy f
(%)
but was excluded in the multivariate analyses because of its statistically significant relationship to dizziness or vertigo. Consequently,
Youden's
indexe (%)
preventive measures are essential not only for patients with dizziantiepileptics or other drugs that may cause dizziness or vertigo.
35
24
ness or vertigo but also for those who are under the influence of
19
3488
Third, mental status or impulsivity is a risk factor item in all three
Negative predictive
valued (%)
tools. Impulsivity in the JHFRAT was found to increase fall-risk by
10 times, consistent with the findings of Harrison et al. (2010); Kim
84.56
changes in mental status or impulsivity. In this study, when a patient
had a change in mental status according to the MFS, their fall-risk
increased 5.9 times. While the HFRM item of “confusion/disorien-
Positive predictive
value (%)c
tation/impulsivity” was a significant variable in the univariate analyses, it was not significant in the multivariate analyses, perhaps due
46.46
risk factor related to patient falls. If the patient has a change in mental status or impulsivity, therefore, an accurate assessment of their
gait is necessary. Concerning the assessment of changes in mental
status and impulsivity, fall-risk was found to be higher when only
74.2
48.51
70.8
64.1
AUC
(%)
36.05
to its being highly related to the Get-Up-and-Go test, a significant
impulsivity was analysed, as in the JHFRAT, compared to the MFS's
Specificity
(%)b
comprehensive factor of mental status or the HFRM's “confusion/
Sensitivity−(1−Specificity) × 100.
(Fallers with high-risk score + Non-fallers with low risk score)/(Total fallers + Total non-fallers) × 100.
57.27
f
in mental status. Moreover, according to Harrison et al. (2010), fall
<.001
<.001
Fallers with high-risk score/total fallers × 100.
Non-fallers with low risk score/total non-fallers × 100.
Fallers with high-risk score/patients with high-risk score × 100.
Non-fallers with low risk score/patients with Low risk score × 100.
b
c
d
295 (53.5)
a
191 (15.4)
256 (46.5)
High-Risk (n = 551)
Low-Risk (n = 1,237)
130 (48.5)
High-risk (n = 268)
Hendrich II fall-risk model (High-risk ≥ 5)
138 (51.5)
1,046 (84.6)
7.41 (5.52, 9.94)
5.66 (3.99, 8.01)
1,203 (79.1)
317 (20.9)
Low-risk (n = 1,520)
Johns Hopkins hospital fall-risk assessment tool (High-risk > 13)
4.86 (3.61, 6.54)
871 (82.7)
470 (63.9)
265 (36.1)
182 (17.3)
Low-risk (n = 1,053)
High-risk (n = 735)
Morse fall scale (High-risk ≥ 45)
the patient is impulsive than when they only experience a change
e
89.71
29.08
64.95
<.001
59.28
Sensitivity
(%)a
P
OR (95% CI)
78.00
disorientation/impulsivity.” This suggests that fall-risk is higher when
Non-faller
(n = 1,341)
Faller
(n = 447)
Characteristics
Comparison of predictive performance of three tools
TA B L E 3
79.15
82.72
et al. (2008) that fall-risk was higher for patients who experienced
injuries are more serious for patients with impulsive tendencies. It
is therefore imperative to add changes in mental status and impulsivity to the fall-risk assessment. Where impulsivity is observed, it is
likely that patients do not move in accordance with their mobility or
comply with nursing interventions proposed to prevent patient falls,
potentially resulting in serious fall injuries. Nursing interventions are
therefore necessary, including focused observation or installation of
fall safety mats that sets off an alarm when a patient leaves their bed.
Fourth, fall history is a risk factor item used in the MFS and
JHFRAT, patients with a history of falling having a 2.2 times higher
fall-risk. This is consistent with studies by Park (2012); Ruchinskas
(2003); and Vassallo et al. (2005). Park (2012) pointed out that patients with a history of falling were 9 times more likely to fall again,
emphasising that those with a fall history or an experience of falling
in the hospital should be categorised in the high-risk group at the
time of admission and be provided with nursing interventions accordingly. Thus, it is necessary to raise awareness of the significance
of accurate fall history reporting in the patient or caregiver interview
process and to review the history of fall-risk assessment in previous
hospital admissions or check the history of unreported falls by physical examination.
This study has the following limitations. First, careful deliberation
is needed in applying the results of this study to other medical environments, as the study specifically targeted inpatients at a tertiary
teaching hospital. Second, this study did not include age and gender
as fall-risk factors. Even though they are known fall-risk factors, age
and gender were used to match the patients between the case and
|
CHO et al.
TA B L E 4
3489
Analysis of relationship between items of three tools
Univariate analysis
Factors
Reference
OR (95% CI)
Multivariable analysis
P
OR (95% CI)
P
AIC
Morse fall scale
History of falling
No
2.19 (1.47, 3.27)
<.001
1.78 (1.09, 2.90)
.02
Secondary disease
No
0.79 (0.56, 1.12)
.19
0.89 (0.60, 1.31)
.56
Intravenous therapy/
Heparin lock
No
1.16 (0.77, 1,74)
.48
1.35 (0.84, 2.16)
.21
Ambulatory aid
None, bed rest, wheel chair
3.09 (2.42, 3.95)
<.001
1.96 (1.47, 2.60)
<.001
Gait
Normal, bed rest, immobile
3.30 (2.66, 4.08)
<.001
2.45 (1.92, 3.12)
<.001
Mental status (Awareness
of own ability)
Oriented to own ability
5.90 (4.13, 8.42)
<.001
4.61 (3.14, 6.77)
<.001
1,005.41
The Johns Hopkins hospital fall-risk assessment tool
Fall history
No
2.19 (1.47, 3.27)
<.001
1.80 (1.08, 3.01)
.03
Elimination
None
3.55 (2.36, 5.33)
<.001
2.89 (1.79, 4.66)
<.001
Number of medications
None
1.94 (1.63, 2.30)
<.001
1.70 (1.35, 2.14)
<.001
Sedated procedure within
past 24 hr
None
0.70 (0.39, 1.26)
.23
1.68 (0.75, 3.75)
.21
Patient care equipment
None
0.90 (0.79, 1.03)
.14
0.76 (0.64, 0.90)
<.01
<.001
10.39 (6.95, 15.54)
<.001
754.82
Medications
Mobility
Requires assistance or
supervision
Normal
14.74 (10.15, 21.39)
Unsteady gait
Normal
4.28 (3.24, 5.66)
<.001
1.96 (1.39, 2.76)
<.001
Visual or auditory
impairment
Normal
5.87 (3.13, 11.00)
<.001
5.52 (2.84, 10.70)
<.001
Altered awareness
Normal
2.81 (1.94, 4.08)
<.001
0.63 (0.35, 1.15)
Impulsive
Normal
9.95 (5.22, 18.95)
<.001
4.46 (1.83, 10.88)
<.01
Lack of understanding
Normal
5.77 (4.06, 8.21)
<.001
2.09 (1.33, 3.30)
<.01
Confusion/disorientation/
impulsivity
No
2.77 (1.92, 3.99)
<.001
1.22 (0.77, 1.94)
.39
Symptomatic depression
No
2.61 (2.05, 3.32)
<.001
1.72 (1.20, 2.47)
<.01
Altered elimination
No
3.88 (2.52, 5.97)
<.001
2.19 (1.31, 3.65)
<.01
Dizziness/vertigo
No
12.04 (8.62, 16.82)
<.001
4.83 (3.30, 7.07)
<.001
Cognition
.14
Hendrich Ⅱ fall-risk model
Antiepileptics
No
1.51 (1.09, 2.09)
.01
1.00 (0.67, 1.50)
.99
Benzodiazepines
No
2.19 (1.62, 2.94)
<.001
1.18 (0.76, 1.84)
.47
“Get up and Go” test
Able to rise in a single
movement
3.34 (2.83, 3.93)
<.001
2.03 (1.68, 2.45)
<.001
814.51
control groups. It is therefore necessary to conduct further studies
from the JHFRAT and HFRM were also collected with reference to
to determine the significance of the relationship of age and gender
the gait and ambulatory aid MFS items and may therefore differ from
to patient falls. Third, the JHFRAT and HFRM risk factor items were
the actual gait assessment. This limitation could be resolved by de-
collected from the ENRs using slightly modified definitions; thus, the
veloping an ENR system that requires more detailed data entry.
findings may differ. For example, the MFS requires that fall occurrences be immediate or within 3 months, whereas the JHFRAT considers occurrences within the past 6 months. However, patient falls
5 | CO N C LU S I O N S
within the past 6 months could not be collected with the ENR data,
which is why the fall history data used in this study were immediate
This study was conducted to identify the tool with the highest pre-
or within 3 months, following the MFS. The gait and mobility data
dictive validity for fall-risk assessment at a tertiary teaching hospital
3490
|
CHO et al.
and to analyse the relationship between the various fall-risk factor
items of the fall-risk assessment tools and fall occurrences. The
comparison of the comprehensive analysis of sensitivity and specificity using the Youden index and AUC showed that the HFRM had
higher predictive validity than the other tools. It is notable that the
ratio of falls among patients in the lowrisk group who experienced
a fall (1-NPV) was also lower for the HFRM, which proved that the
validity of the HFRM is the best in the lowrisk group who is overlooked in the fall intervention. Finally, each risk factor item in the
three tools was analysed as to whether it was a valid predictor of
fall-risk. It was found that gait, dizziness or vertigo, mental status
or impulsivity, and fall history were the most significant fall-risk
factors.
6 | R E LE VA N C E TO C LI N I C A L PR AC TI C E
This study is meaningful in that it was conducted in an acute care
setting hospital in which patients with various characteristics are
admitted. In addition, this study identified significant fall-risk factors, regardless of fall-risk assessment tools. Therefore, whatever
the fall-risk assessment tool used in acute care setting hospital, we
recommend to consider gait, dizziness or vertigo, mental status or
impulsivity, and fall history which are identified as significant risk
factors in this study.
C O N FL I C T O F I N T E R E S T
The authors have no conflict of interest to declare.
AU T H O R C O N T R I B U T I O N
Eun Hee Cho and Hyeoun-Ae Park contributed to conception and
design of the study in discussion with all authors. Eun-Hee Cho
took the lead in extracting and analysing the data, and writing the
manuscript with the support of Yung Jung Woo. Hyeoun-Ae Park
supervised analysis and interpretation of data, manuscript writing,
submission and revision. Eun Hee Cho, Yun Jung Woo, Arum Han,
and Yoo Chung Chung extracted the data from EHRs under the supervision of Yeon Hee Kim. All authors provided critical feedback
and helped shape the research, analysis and manuscript.
ORCID
https://orcid.org/0000-0003-4022-6445
Eun Hee Cho
https://orcid.org/0000-0002-9363-4752
Yun Jung Woo
Arum Han
https://orcid.org/0000-0001-6934-7131
Yoon Chung Chung
Yeon Hee Kim
Hyeoun-Ae Park
https://orcid.org/0000-0002-5210-1052
https://orcid.org/0000-0002-9409-7354
https://orcid.org/0000-0002-3770-4998
REFERENCES
Ambrose, A. F., Paul, G., & Hausdorff, J. M. (2013). Risk factors for falls
among older adults: A review of the literature. Maturitas, 75(1), 51–
61. https://doi.org/10.1016/j.matur​itas.2013.02.009
Aranda-Gallardo, M., Morales-Asencio, J. M., Canca-Sanchez, J. C.,
Barrero-Sojo, S., Perez-Jimenez, C., Morales-Fernandez, A., …
Mora-Banderas, A. M. (2013). Instruments for assessing the risk
of falls in acute hospitalized patients: A systematic review and
meta-analysis. BMC Health Services Research, 13, 122. https://doi.
org/10.1186/1472-6963-13-122
Baek, S., Piao, J., Jin, Y., & Lee, S. M. (2014). Validity of the Morse Fall
Scale implemented in an electronic medical record system. Journal
of Clinical Nursing, 23(17–18), 2434–2440. https://doi.org/10.1111/
jocn.12359
Caldevilla, M. N., Costa, M. A., Teles, P., & Ferreira, P. M. (2013).
Evaluation and cross-cultural adaptation of the Hendrich II fall risk
model to Portuguese. Scandinavian Journal of Caring Sciences, 27(2),
468–474. https://doi.org/10.1111/j.1471-6712.2012.01031.x
Callis, N. (2016). Falls prevention: Identification of predictive fall risk factors. Applied Nursing Research, 29, 53–58. https://doi.org/10.1016/j.
apnr.2015.05.007
Dunne, T. J., Gaboury, I., & Ashe, M. C. (2014). Falls in hospital increase length of stay regardless of degree of harm. Journal
of Evaluation in Clinical Practice, 20(4), 396–400. https://doi.
org/10.1111/jep.12144
Eagles, D., Yadav, K., Perry, J. J., Sirois, M. J., & Emond, M. (2018).
Mobility assessments of geriatric emergency department patients:
A systematic review. CJEM, 20(3), 353–361. https://doi.org/10.1017/
cem.2017.46
Grimes, D. A., & Schulz, K. F. (2005). Compared to what? Finding controls
for case-control studies. The Lancet, 365(9468), 1429–1433. https://
doi.org/10.1016/S0140​-6736(05)66379​-9
Han, J., Xu, L., Zhou, C., Wang, J., Li, J., Hao, X., … Yang, N. (2017). Stratify,
Hendrich II fall risk model and Morse fall scale used in predicting
the risk of falling for elderly in-patients. Biomedical Research, Special
Issue, S439–S442.
Harrison, B. E., Ferrari, M., Campbell, C., Maddens, M., & Whall, A. L.
(2010). Evaluating the relationship between inattention and impulsivity-related falls in hospitalized older adults. Geriatric Nursing, 31(1),
8–16. https://doi.org/10.1016/j.gerin​urse.2009.09.002
Hendrich, A. L., Bender, P. S., & Nyhuis, A. (2003). Validation of the
Hendrich II Fall Risk Model: A large concurrent case/control study of
hospitalized patients. Applied Nursing Research, 16(1), 9–21. https://
doi.org/10.1053/apnr.2003.YAPNR2
Hong, H. J., Kim, N. C., Jin, Y., Piao, J., & Lee, S. M. (2015). Trigger factors
and outcomes of falls among korean hospitalized patients: Analysis
of electronic medical records. Clinical Nursing Research, 24(1), 51–72.
https://doi.org/10.1177/10547​73814​524225
Hou, W. H., Kang, C. M., Ho, M. H., Kuo, J. M., Chen, H. L., & Chang, W.
Y. (2017). Evaluation of an inpatient fall risk screening tool to identify the most critical fall risk factors in inpatients. Journal of Clinical
Nursing, 26(5–6), 698–706. https://doi.org/10.1111/jocn.13510
Ivziku, D., Matarese, M., & Pedone, C. (2011). Predictive validity of the
Hendrich fall risk model II in an acute geriatric unit. International
Journal of Nursing Studies, 48(4), 468–474. https://doi.org/10.1016/j.
ijnur​stu.2010.09.002
Jang, I.-S., & Lee, S.-G. (2014). Fall risk factors and characteristics of an
acute hospital setting across clinical departments. Journal of Korean
Academy of Fundamentals of Nursing, 21(3), 264–274. https://doi.
org/10.7739/jkafn.2014.21.3.264
Jung, H., & Park, H. A. (2018). Testing the predictive validity of the
Hendrich II Fall Risk Model. Western Journal of Nursing Research,
40(12), 1785–1799. https://doi.org/10.1177/01939​45918​766554
Kim, E.-K., Lee, J. C., & Eom, M.-R. (2008). Falls risk factors of inpatients.
Journal of Korean Academy of Nursing, 38(5), 676–684. https://doi.
org/10.4040/jkan.2008.38.5.676
Kim, E. A., Mordiffi, S. Z., Bee, W. H., Devi, K., & Evans, D. (2007).
Evaluation of three fall-risk assessment tools in an acute care setting. Journal of Advanced Nursing, 60(4), 427–435. https://doi.
org/10.1111/j.1365-2648.2007.04419.x
Kim, Y. S., & Choi-Kwon, S. (2013). Fall risk factors and fall risk assessment of inpatients. Korean Journal of Adult Nursing, 25(1), 74–82.
https://doi.org/10.7475/kjan.2013.25.1.74
|
CHO et al.
Morse, J. M., Black, C., Oberle, K., & Donahue, P. (1989). A prospective
study to identify the fall-prone patient. Social Science and Medicine,
28(1), 81–86. https://doi.org/10.1016/0277-9536(89)90309​-2
Park, I. S. (2012). Predictive effects of previous fall history on accuracy
of fall risk assessment tool in acute care settings. Journal of Korean
Academy of Fundamentals of Nursing, 19(4), 444–452. https://doi.
org/10.7739/jkafn.2012.19.4.444
Poe, S. S., Cvach, M. M., Gartrelu, D. G., Radzik, B. R., & Joy, T. L. (2005).
An evidence-based approach to fall risk assessment, prevention, and
management: Lessons learned. Journal of Nursing Care Quality, 20(2),
107–116; quiz 117–108. https://doi.org/10.1097/00001​786-20050​
4000-00004
Poe, S. S., Dawson, P. B., Cvach, M., Burnett, M., Kumble, S., Lewis, M., …
Hill, E. E. (2018). The Johns Hopkins fall risk assessment tool: A study
of reliability and validity. Journal of Nursing Care Quality, 33(1), 10–19.
https://doi.org/10.1097/NCQ.00000​0 0000​0 00301
Ruchinskas, R. (2003). Clinical prediction of falls in the elderly. American
Journal of Physical Medicine & Rehabilitation, 82(4), 273–278. https://
doi.org/10.1097/01.PHM.00000​56990.35007.C8
Vassallo, M., Stockdale, R., Sharma, J. C., Briggs, R., & Allen, S. (2005).
A comparative study of the use of four fall risk assessment tools on
3491
acute medical wards. Journal of the American Geriatrics Society, 53(6),
1034–1038. https://doi.org/10.1111/j.1532-5415.2005.53316.x
Yang, H.-M., & Chun, B.-C. (2009). Falls in the general hospital inpatients:
Incidence, associated factors. Quality Improvement in Health Care,
15(2), 107–120.
S U P P O R T I N G I N FO R M AT I O N
Additional supporting information may be found online in the
Supporting Information section.
How to cite this article: Cho EH, Woo YJ, Han A, Chung YC,
Kim YH, Park H-A. Comparison of the predictive validity of
three fall risk assessment tools and analysis of fall-risk factors
at a tertiary teaching hospital. J Clin Nurs. 2020;29:3482–
3493. https://doi.org/10.1111/jocn.15387
3492
|
CHO et al.
APPENDIX 1
T H E M O R S E FA L L S C A L E ( M F S )
Risk factor
Point
1. History of falling
Yes
25
No
0
Yes
15
No
0
Furniture
30
Crutch/cane/walker
15
None/bed rest/nurse assist
0
Yes
20
No
0
Impaired
20
Weak
10
Normal/bed rest/wheel chair
0
Overestimates/forgets limitations
15
Oriented to own ability
0
2. Secondary diagnosis
3. Ambulatory aid
4. Intravenous therapy/injection cap
5. Gait
6. Mental status
high-risk ≥ 45
APPENDIX 2
T H E J O H N S H O P K I N S H O S P I TA L FA L L R I S K A S S E S S M E N T TO O L ( J FR AT )
Risk factor
Point
1. Age (single-select)
60–69 years
1
70–79 years
2
greater than or equal to 80 years
3
2. Fall history (single-select)
One fall within 6 months before admission
5
3. Elimination, bowel and urine (single-select)
Incontinence
2
Urgency or frequency
2
Urgency/frequency and incontinence
4
4. Medications: Includes PCA/opiates, anticonvulsants, antihypertensives, diuretics, hypnotics, laxatives, sedatives, and
psychotropics (single-select)
On 1 high fall risk drug
3
On 2 or more high fall risk drugs
5
Sedated procedure within past 24 hr
7
5. Patient Care Equipment: Any equipment that tethers patient
(e.g. IV infusion, chest tube, indwelling catheter, SCDs, etc.)
(single-select)
One present
1
Two presents
2
3 or more present
3
6. Mobility (multi-select; choose all that apply and add points
together)
Requires assistance or supervision for mobility, transfer, or
ambulation
2
Unsteady gait
2
7. Cognition (multi-select; choose all that apply and add points
together)
high-risk > 13
Visual or auditory impairment affecting mobility
2
Altered awareness of immediate physical environment
1
Impulsive
2
Lack of understanding of one's physical and cognitive limitations
3
|
CHO et al.
APPENDIX 3
T H E H E N D R I C H I I FA L L R I S K M O D E L ( H F R M )
Risk factor
Point
1. Confuse/disorientation/impulsivity
Yes
4
No
0
Yes
2
No
0
Yes
1
No
0
Yes
1
2. Symptomatic depression
3. Altered elimination
4. Dizziness/vertigo
No
0
5. Gender (Male)
Male
1
Female
0
6. Any administered antiepileptics; Carbamazepine, Divalproex Sodium,
Ethosuximide, Fosphenytoin, Gabapentin, Lamotrigine, Phenobarbital,
Phenytoin, Primidone, Topiramate, Valproic Acid etc.
Yes
2
No
0
7. Any administered benzodiazepines; Alprazolam, Clonazepam, Diazepam,
Lorazepam, Midazolam, Triazolam etc.
Yes
1
No
0
8. “Get up and go” test
Ability to rise in single movement – No loss of balance
with steps
0
Pushes up, successful in one attempt
1
Multiple attempts but successful
3
Unable to rise without assistance during test
4
high-risk ≥ 5
3493