Effectiveness of Fall Prevention Programs in Nursing Homes: A Systematic Review and Meta-Analysis

CLINICAL INVESTIGATIONS
Characteristics and Effectiveness of Fall Prevention Programs in
Nursing Homes: A Systematic Review and Meta-Analysis of
Randomized Controlled Trials
Ellen Vlaeyen, MSN, a Joke Coussement, MSN, a,b Greet Leysens, MSN, a Elisa Van der Elst, MSN, a
Kim Delbaere, MPT, PhD, c Dirk Cambier, MPT, PhD, d Kris Denhaerynck, MSN, PhD, e
Stefan Goemaere, MD, f Arlette Wertelaers, MD, g Fabienne Dobbels, PhD, a
Eddy Dejaeger, MD, PhD, h and Koen Milisen, MSN, PhD, a,h on behalf of the Center of Expertise
for Fall and Fracture Prevention Flanders
OBJECTIVES: To determine characteristics and effective-
ness of prevention programs on fall-related outcomes in a
defined setting.
DESIGN: Systematic review and meta-analysis.
SETTING: A clearly described subgroup of nursing homes
defined as residential facilities that provide 24-hour-a-day
surveillance, personal care, and limited clinical care for
persons who are typically elderly and infirm.
PARTICIPANTS: Nursing home residents (N = 22,915).
MEASUREMENTS: The primary outcomes were number
of falls, fallers, and recurrent fallers.
RESULTS: Thirteen studies met the inclusion criteria.
Six fall prevention programs were single (one interven-
tion component provided to the residents), one was mul-
tiple (two or more intervention components not
customized to individual fall risk), and six were multi-
factorial (two or more intervention components custom-
ized to each resident’s fall risk). Meta-analysis found
significantly fewer recurrent fallers in the intervention
groups (4 studies, relative risk (RR) = 0.79, 95% confi-
dence interval (CI) = 0.65–0.97) but no significant effect
of the intervention on fallers (6 studies, RR = 0.97,
95% CI = 0.84–1.11) or falls (10 studies, RR = 0.93,
95% CI = 0.76–1.13). Multifactorial interventions signifi-
cantly reduced falls (4 studies, RR = 0.67, 95%
CI = 0.55–0.82) and the number of recurrent fallers (4
studies, RR = 0.79, CI = 0.65–0.97), whereas single or
multiple interventions did not. Training and education
showed a significant harmful effect in the intervention
groups on the number of falls (2 studies, RR = 1.29,
95% CI = 1.23–1.36).
CONCLUSION: This meta-analysis failed to reveal a
significant effect of fall prevention interventions on falls or
fallers but, for the first time, showed that fall prevention
interventions significantly reduced the number of recurrent
fallers by 21%. J Am Geriatr Soc 63:211–221, 2015.
Key words: accidental falls; prevention; multifactorial
interventions; residential aged care facilities; meta-
analysis
Nursing home residents have a high risk of falling. The
average fall incidence is estimated to be 1.6 falls per
bed per year, with almost half of residents falling more
than once a year.1–3 Falls in nursing homes often lead to
serious injuries, with for example, an estimated hip frac-
ture incidence rate of 4% annually. 3–5 Previous studies
showed that, within 1 year after a fall-related hip fracture,
12% of residents incur a new fracture, and 31% die as a
result. 5,6 Apart from the physical burden, falls often have
psychological consequences such as fear of falling and
poor quality of life. Falls also have a significant economic
burden.7,8
The Prevention of Falls Network Europe (ProFaNE)
taxonomy divides fall prevention programs into three types:
single programs, which include one intervention component
From the a
Department of Public Health and Primary Care, Health Services
and Nursing Research, KU Leuven, Leuven, Belgium; b
Vzw Rusthuizen
Zusters van Berlaar, Berlaar, Belgium; c
Neuroscience Research Australia,
University of New South Wales, Sydney, Australia; d
Rehabilitation
Sciences and Physiotherapy, Ghent University, Ghent, Belgium; e
Institute
of Nursing Science, University of Basel, Basel, Switzerland; f Unit for
Osteoporosis and Metabolic Bone Disease, Department of Rheumatology
and Endocrinology, Ghent University Hospital, Ghent, Belgium; g
Domus
Medica, Society of Flemish General Practitioners, Antwerpen, Belgium;
and h
Division of Geriatric Medicine, University Hospitals Leuven, Leuven,
Belgium.
Address correspondence to Koen Milisen, KU Leuven, Department of
Public Health and Primary Care, Health Services and Nursing Research,
Kapucijnenvoer 35, 4th floor, P.B. 7001, Leuven 3000, Belgium.
E-mail: [e-mail]
DOI: 10.1111/jgs.13254
JAGS 63:211–221, 2015
© 2015, Copyright the Authors
Journal compilation © 2015, The American Geriatrics Society 0002-8614/15/$15.00

provided to all residents (e.g., supervised exercises); multiple
programs, which include two or more intervention compo-
nents provided to all residents (e.g., supervised exercise and
staff training); and multifactorial programs, which include
two or more customized intervention components that tar-
get each resident’s fall risk profile.9 No conclusive evidence
exists on the effectiveness of fall prevention programs in
nursing homes, partly because of differing study
approaches.10–16 For example, five of seven published
reviews used a narrative approach,10–14 one of which
reviewed only multifactorial interventions. 13 Two recent
systematic reviews 15,16 used a meta-analytical method and
reported no effect of any type of intervention, with the
exception of one single intervention that showed improve-
ment in number of falls and fallers after supplementing resi-
dents’ diets with vitamin D.15
These reviews did not distinguish between fallers and
recurrent fallers. This is a missed opportunity, given the
high frequency of recurrent fallers and their influence on
the total number of falls. Furthermore, previous reviews
compared studies that used heterogeneous groups of resi-
dents from various care settings that had major differences
in care intensity or used vague terminology to define the
care settings (e.g., residential or nursing care facilities, 13,15
nursing homes,10–12 care homes,16 and long-term care
facilities 14
). This could partly explain why evidence on
effective fall prevention strategies in nursing homes is less
conclusive than reviews specifically focusing on a more-
clear-cut delineated population of community-dwelling
elderly adults.17
The current review aimed to determine the characteris-
tics and effectiveness of single, multiple, and multifactorial
fall prevention programs on the number of falls, fallers,
and recurrent fallers in older persons who permanently
reside in a nursing home. A nursing home was defined as
“a residential facility that provides 24-hour-a-day surveil-
lance, personal care, and limited clinical care for persons
who are typically elderly and infirm” and excluded post-
hospital skilled nursing care, rehabilitation, and long-term
care for younger people with illnesses, injuries, functional
disabilities, or cognitive impairment.7
METHODS
The review protocol was registered on PROSPERO
(CRD42011001687) 18 and conducted in concordance with
PRISMA guidelines.19,20
Search Strategy
A systematic literature search was conducted in multiple
databases (MEDLINE, EMBASE, Cochrane Central Regis-
ter of Controlled Trials, PEDro, CINAHL, SportDiscus),
restricting the search to articles published from database
inception to September 2013. Depending on the selected
database, Medical Subject Headings, a thesaurus, or free
text was combined with the Boolean operators “AND/
OR” to build a search strategy. Search terms were “acci-
dental falls,” “falls,” “faller*,” “aged,” “older,” “elderly,”
“nursing homes,” “residential facilities,” “long-term care,”
“institutionalization,” “residential*,” and “prevention and
control.”
Relevant studies were identified using three steps.
First, two independent reviewers (EV, JC) conducted an
initial study selection based on title and abstract. Second,
three reviewers (EV, JC, KM) obtained and examined full-
text copies of all articles meeting initial selection criteria.
Disagreement was resolved through discussion with three
additional reviewers (GL, EVdE, ED). Third, reference lists
of articles meeting the inclusion criteria were screened for
additional relevant papers.
Inclusion Criteria
Studies had to meet the following criteria.
Setting
The study had to be conducted in a nursing home, as
defined previously; 7 other kinds of residential care facilities
were excluded. If the setting was in doubt, an attempt was
made to contact the authors for clarification. When studies
included nursing homes and other facilities (e.g., assisted
living facilities), the study was included only when sepa-
rate results for the nursing home population were available
in the article.
Design
The study had to be an original or a priori secondary
analysis of individual-level or cluster randomized con-
trolled trials (RCT).
Objective
The intervention had to include single, multiple, or multi-
factorial fall prevention programs designed to prevent
falls.
Outcomes
The study had to examine intervention effect on number
of falls, fallers, or recurrent fallers. The term “faller”
included all residents sustaining at least one fall during the
intervention or follow-up period. In the same way, recur-
rent fallers were defined as residents sustaining two or
more falls.
Duration
The duration from the start of the intervention (including
follow-up) had to be 6 months or longer.
Language
Only publications in English, French, German, or Dutch
were considered.
Risk of Bias Assessment
The methodological quality of each study was assessed
using the Cochrane methodological quality assessment
scheme (Table 1). 21 One loss-to-follow-up criterion was
added and evaluated according to the question: “Were the
majority of participants still in the sample at the end of
the study?” Majority was operationally defined as 80% or
higher.
212 VLAEYEN ET AL. FEBRUARY 2015–VOL. 63, NO. 2 JAGS

Table 1. Methodological Quality
Study
Fall Incident
Clearly
Defined and
Related to
Staff
Inclusion and
Exclusion
Criteria
Clearly
Defined
Blinded
Randomization
Treatment and
Control Groups
Comparable at
Baseline
Identical
Standard Care
Programs for
Both Groups
Blinded
Treatment
Providers
Blinded
Outcome
Assessors
Blinded
Subjects
Identical
Ascertainment
of Outcomes
Intention-
to-Treat
Analysis
Loss
to
Follow-
Upa
Total
Score
(0–22)
Becker 34b 1 2 2 1 0 0 0 0 2 2 2 12
Cox 27 0 1 2 0 0 0 0 0 2 0 1 6
Dyer 36 0 2 2 1 0 0 0 0 2 2 2 11
Kerse 37 2 2 2 2 0 0 2 0 2 2 0 14
Lapane 29 1 2 0 1 0 0 0 0 2 0 0 6
Law 31 0 2 2 2 0 0 1 0 2 2 0 11
McMurdo 38 1 2 1 2 0 0 2 0 2 2 0 12
Neyens 39 0 2 2 2 0 0 0 0 2 2 0 10
Patterson 30 0 2 2 1 0 0 1 0 2 1 0 9
Rapp 35b 1 2 2 2 0 0 0 0 2 0 0 9
Ray 40 1 2 2 2 0 0 2 0 2 2 0 13
Schnelle 33 1 2 2 2 0 0 2 0 2 0 0 11
Schoenfelder 32 1 2 0 1 0 0 0 0 2 0 2 8
Ward 28 0 0 2 1 0 0 0 0 2 2 0 7
Scores: 0 = not meeting the criterion, mentioned but unclear, or not mentioned; 1 = partially meeting the criterion; 2 = completely meeting the criterion.
a
Were the majority (≥80%) of participants still in the sample at the end of the study? Loss-to-follow-up was defined as participants who died, became terminally ill, or moved out of the nursing home during the
study or follow-up period.
b
Same intervention described in both articles.
JAGS FEBRUARY 2015–VOL. 63, NO. 2 EFFECT OF FALL PREVENTION IN NURSING HOMES 213

Five reviewers (EV, JC, KD, ED, KM) independently
scored methodological quality on a scale ranging from 0
to 2, depending on whether the criterion was not fulfilled
or not mentioned (0), partially met (1), or completely met
(2). The total score ranged between 0 and 22, with higher
scores indicating better quality. Disagreement was resolved
through discussion and consensus.
Data Synthesis and Analysis
Data were synthesized to collate and summarize individual
study results. One reviewer (EV) collected data on charac-
teristics and effectiveness of the fall prevention programs.
Two reviewers (GL, EV & E) independently confirmed the
accuracy of the data synthesis. The ProFaNE taxonomy
for fall prevention interventions was used to classify and
describe the intervention programs.9 The programs were
categorized according to the ProFaNE taxonomy: interven-
tion type (single, multiple, or multifactorial) and interven-
tion component (e.g., exercise, medication).9 Effectiveness
was classified according to outcome. For number of falls,
the relative risk of falls per resident years was calculated
(ratio of the number of falls per resident year in the inter-
vention group to number of falls per resident year in the
control group). For number of fallers and recurrent fallers,
the relative risk was calculated (ratio of proportion of
(recurrent) fallers in intervention group to corresponding
proportion in control group).
Effect size pooling was followed by regression analyses
using a random effects approach. 22 Publication bias was
checked using the trim and fill method, which alerts for
publication bias if more than three studies on the funnel
plot are estimated to be concealed (assuming b > 0.80 and
a = 0.05). 23,24 Heterogeneity of variance tests were
used to examine whether between-study variability was
significantly different from 0 using weighted error sum of
squares Q. 25 Calculated I² parameters reflected the propor-
tion of between-study variability to total variability.26
An a priori subgroup analysis was conducted accord-
ing to ProFaNE intervention type (single, multiple, multi-
factorial) and to determine any possible effects the
intervention might have had on residents with dementia.
For the latter analysis, dementia was converted into a cate-
gorical variable, based on the methodology that Oliver
and colleagues used.16 Studies were categorized into three
groups ranging from 1 to 3 according to the prevalence of
dementia, with 1 being assigned to studies with less than
40% prevalence, 2 being assigned to studies with a preva-
lence between 40% to 69%, and 3 being assigned to stud-
ies with 70% or more of residents with dementia being
included in the study. Studies with an unspecified number
of participants with dementia were excluded from the sub-
group analysis. Regression analyses were conducted to
determine whether dementia prevalence predicted effect
size in this subgroup analysis.
RESULTS
Selected Studies
Figure 1 summarizes the results of the different steps for
identifying appropriate articles for review. 19,20 Fourteen
articles, describing 13 studies, met the inclusion criteria.27–40
Two studies 34,35 were combined as one study in the meta-
analysis because they reported data on the same cohort of
individuals.
Risk of Bias Assessment
Overall quality scores ranged from 6 to 14 (Table 1). The
definition of a fall was reported in eight studies, 29,32–
35,37,38,40 but in only one study was the definition clearly
explained to the staff collecting and reporting the data. 37
Inclusion and exclusion criteria were clearly defined in all
but two studies. 27,28 None of the included studies had
identical standard care programs for both intervention and
control groups, neither did they employ treatment blinding
for the treatment providers or participants. Outcome asses-
sors were blinded in four studies. 33,37,38,40 Eight studies
used intention-to-treat analysis. 28,31,34,36–40 The majority
of participants were still in the sample at the end of the
study in three studies. 32,34,36
Study Characteristics
Seven studies were conducted in Europe, 27,30,31,34–36,38,39
four of which were in the United Kingdom.27,31,36,38 Four
were conducted in the United States29,32,33,40 and two in
Australia/Oceania28,37 (Table 2). Follow-up ranged from
6 to 17 months. Two studies were categorized as being
individual RCTs, 32,33 and 12 were cluster RCTs.27–31,34–40
With regard to ProFaNE intervention type, there were six
single,27–32 one multiple,33 and six multifactorial34–40 fall
prevention programs (Table 3). The included articles had
22,915 residents, with an overall mean age range of 82
to 88.
Overall Effects on Number of Falls, Fallers, and
Recurrent Fallers
The number of falls was reported in 12 studies
(Table 4). 27–36,38,39 In nine studies, the intervention and
control groups did not differ significantly in number of
falls.27–33,36,38 Two multifactorial studies 34,39 showed a
significant decrease in falls of 36% and 45%, respectively,
over a 12-month period for the intervention groups. An a
priori secondary subgroup analysis 35 of one study 34
showed that the effect was only significant in cognitively
impaired residents (relative risk (RR) = 0.49, 95% confi-
dence interval (CI) = 0.35–0.69) and not in cognitively
intact residents (hazard rate (HR) = 0.91, 95% CI = 0.68–
1.22). Pooled data from 10 studies showed no effect
on number of falls (RR = 0.93, 95% CI = 0.76–1.13;
I² = 89.8%, P < .001). 27–31,33,34,36,38,39
Seven studies assessed the number of fallers as an out-
come measure,31,33–38 of which two multifactorial studies
showed significantly fewer fallers in the intervention group
than in the control group. These two studies reported
25%34 and 30%35 fewer fallers, although the pooled risk
estimates failed to demonstrate a beneficial effect of
the intervention on number of fallers (RR = 0.97, 95%
CI = 0.84–1.11; I² = 47.9%, P = .09).31,33,34,36–38
The number of recurrent fallers was assessed in four
multifactorial studies. 34,36,38,40 One study34 described a
214 VLAEYEN ET AL. FEBRUARY 2015–VOL. 63, NO. 2 JAGS