JAMDA 13 (2012) 188.e13e188.

e21

JAMDA
journal homepage: www.jamda.com

Original Study

Effectiveness of Intervention Programs In Preventing Falls: A Systematic Review of

Recent 10 Years and Meta-Analysis
Myunghan Choi PhD, MPH, APRN-BC a, b, *, Melvin Hector MD b, c
a
College of Nursing & Health Innovation, Arizona State University, Phoenix, AZ
b
IPC The Hospitalist Company, Tucson, AZ
c
College of Medicine, University of Arizona, Tucson, AZ

a b s t r a c t
Keywords: Objective: To examine the reported effectiveness of fall-prevention programs for older adults by
Falls reviewing randomized controlled trials from 2000 to 2009.
prevention programs
Design: Systematic review and meta-analysis of randomized controlled trials.
meta-analysis
older adults
Data Sources: A systematic literature search of articles was conducted using 5 electronic databases
(Medline, PubMed, PsycINFO, CINAHL, and RefWorks), including articles describing interventions
designed to prevent falls, in English with full text availability, from 2000 through 2009.
Review Methods: Of a potential 227 studies, we identified 17 randomized controlled trials with a duration
of intervention of at least 5 months of follow-up. Inclusion and exclusion criteria were used to assess the
methodological qualities of the studies. We excluded unidentified study design, quasi-experimental
studies, and/or studies that were nonspecific regarding inclusion criteria.
Data Extraction: Primary outcome measures were number of falls and fall rate. Methodological quality
assessment included internal and external validity, reporting, and power. Data were extracted inde-
pendently by 2 investigators and analyzed using a random-effects model. We analyzed the effectiveness
of these fall intervention programs using their risk ratios (RR) in 2 single-intervention versus 15
multifactorial intervention trials, 3 nursing homes versus 14 community randomized controlled trials,
and 8 Model 1 (initial intervention with subsequent follow up) versus 9 Model II (ongoing intervention
throughout the follow-up) studies.
Results: The combined RR for the number of falls among 17 studies was 0.855 (z ¼ e2.168; p ¼ .030; 95%
CI ¼ 0.742e0.985; Q ¼ 196.204, df ¼ 16, P ¼ .000, I2 ¼ 91.845), demonstrating that fall-prevention
programs across the studies were effective by reducing fall rates by 14%, but with substantial hetero-
geneity. Subgroup analysis indicated that there was a significant fall reduction of 14% in multifactorial
intervention (RR ¼ 0.856, z ¼ e2.039, P ¼ .041) with no variation between multifactorial and single-
intervention groups (Q ¼ 0.002, P ¼ .961), 55% in the nursing home setting (RR ¼ 0.453, z ¼ e9.366,
P ¼ .000) with significant variation between nursing home and community groups (Q ¼ 62.788, P ¼
.000), and no significant effect was gained by dividing studies into either in Model I or II. Sensitivity
analysis found homogeneity (Q ¼ 18.582, df ¼ 12, P ¼ .099, I2 ¼ 35.423) across studies with a 9% overall
fall reduction (RR ¼ 0.906, 95% CI ¼ 0.853e0.963, z ¼ e3.179, P ¼ .001), including a fall-reduction rate of
10% in multifactorial intervention (RR ¼ 0.904, z ¼ e3.036, P ¼ .002), 9% in community (RR ¼ 0.909, z ¼
e3.179, P ¼ .001), and 12% in Model I (RR ¼ 0.876, z ¼ e3.534, P ¼ .000) with no variations among all the
groups. Meta regression suggested that the model fit explained 68.6% of the relevant variance.
Conclusions: The meta-sensitivity analysis indicates that randomized controlled trials of fall-prevention
programs conducted within the past 10 years (2000e2009) are effective in overall reduction of fall
rates of 9% with a reduction of fall rates of 10% in multifactorial interventions, 9% in community settings,
and 12% in Model I interventions (initial intervention efforts and then subsequent follow-up).
Published by Elsevier Inc. on behalf of the American Medical Directors Association, Inc.

Falls are the leading cause of injury deaths among older adults
The authors declare no conflicts of interest. and the leading cause of hospital admissions for trauma in this age
* Address correspondence to Myunghan Choi, PhD, MPH, APRN-BC, College of
group.1,2 Direct medical cost estimates totaled $19.2 billion for fatal
Nursing & Health Innovation, Arizona State University, 500 N. Third Street, Phoenix,
AZ 85004. and nonfatal injuries in the United States in 2000.3,4 Earlier studies
E-mail address: Myunghan.choi@asu.edu (M. Choi). have suggested that fall-related injuries could be prevented with
1525-8610/$ - see front matter Published by Elsevier Inc. on behalf of the American Medical Directors Association, Inc.
doi:10.1016/j.jamda.2011.04.022
 M. Choi, M. Hector / JAMDA 13 (2012) 188.e13e188.e21
either multifactorial or individualized intervention.5,6 In 2003,
Gillespie et al6 reported that multifactorial-intervention programs
and single-intervention programs of muscle strengthening and
balance retraining were likely to be effective. Gates et al’s5 meta-
analysis of multifactorial interventions in 2008 found little
evidence in support.
Previous research has identified that fall-intervention programs
were effective when focusing on a multidisciplinary team inter-
vention.7e10,19 Single-intervention programs were effective when
focusing on a certain exercise program, when implementing
a specific home safety assessment, providing a high dose of vitamin
D supplementation, and so forth.11e15 Community intervention
programs were effective when focusing on a comprehensive at-
home assessment, individualized risk reductions, and/or educa-
tion programs used to prevent falls.10,16e20 In 2004, Chang et al7
concluded that a multifactorial fall-risk assessment and manage-
ment program was the most effective intervention, and that a single
intervention program of exercise also had a beneficial effect.
Despite these and many other intervention programs, fall-
related injuries among older adults in general seem to have
further decreased only slightly. Several meta-analysis studies based
on fall-related studies from the 1990s and 2000s including
randomized controlled trials (RCTs) and quasi-experimental studies
reported disparate conclusions.5,7
Purpose
Several meta-analysis studies reported the effectiveness of fall-
prevention programs focusing on the previous decade (1990s to
early 2000) for older adults.7,8,12,20e23 Since then, a variety of fall-
intervention programs have been developed and implemented
either in communities or nursing homes for frail older adults.
However, there is no recent meta-analysis indicating that newly
developed intervention programs are any more or less reliably
effective in preventing falls for older adults. Our analysis was an
attempt to derive scientific evidence of the effectiveness of fall-
prevention programs and to make recommendations to health
care providers about the effectiveness of fall-intervention
programs.
Methods
Identification of Studies
A thorough search of the scientific and medical literature was
conducted using major biomedical electronic databases: Medline,
PubMed, PsycINFO, CINAHL, and RefWorks. This rigorous literature
examination selected articles in peer-reviewed journals published
in the English language, with full-text availability, targeting both
men and women, in RCTs, with primary outcome measures as
either the number of falls or the fall rate, with fall-intervention
follow-up of a minimum of 5 months, and published from 2000
to 2009. We excluded articles with unidentified study design,
cohort studies, case-control studies, quasi-experimental studies,
quasi-randomized controlled studies, and community studies
conducted in environments where essential postintervention
follow-up was impractical.24
There were 33 studies duplicated across databases that were
excluded. The key words used were the combination of falls,
recurrent falls, fall-prevention programs, interventions, programs,
injuries, older adults, RCTs, and/or long term care facilities. A total
of 17,325 studies were generated from the 5 electronic databases.
Studies with titles and abstracts unlikely to be relevant were
excluded (n ¼ 15,419) followed by exclusion of studies that did not
identify an RCT (n ¼ 1619). A total of 287 potentially relevant
188.e14
studies were reviewed by the primary author (M.C.). We excluded
studies of fall-prevention programs for children (n ¼ 27). The
primary search (completed from July 2009 to March 2010) gener-
ated a total of 227 studies: 22 studies from CINAHL, 62 studies from
Medline, 37 studies from PsycInfo, 64 studies from PubMed, and 42
from RefWorks. Among these, we excluded studies that did not
report fall-intervention programs or effectiveness of the program
(n ¼ 151), those that identified unclear intervention follow-up or
were quasi-experimental in design (n ¼ 40), those that included
men or women only (n ¼ 15), or those that were studies where
postintervention follow-up was impractical (n ¼ 4). Of the 227
publications that met the criteria for meta-analysis, 17 studies
remained (Figure 1).
Quality Assessment
Each of these 17 studies was classified independently using the
Downs and Black Checklist for Measuring Study Quality (DB
checklist).25 The DB checklist included a total of 27 items including
five subcategories: 10 items for reporting, three items for external
validity, seven items for internal validity, six items for selection
bias, and one item for power. All items were measured one for “Yes”
or 0 for “No” or “Unable to determine.” Inter-rater reliabilities were
assessed to examine the extent of consensus on use of the scale
between two completely blinded raters.25 After the completion of
the quality assessment by each individual, the two raters presented
their own quality assessment. Because the data are categorical and
raters are two, the consensus was examined using Cohen’s Kappa as
an inter-rater reliability measure.26 The Cohen’s Kappa was calcu-
lated using a 3  3 table and number of agreement in five categories
on the 17 studies. Five categories included reporting (n ¼ 10),
external validity (n ¼ 3), internal validity (n ¼ 7), selection bias (n ¼
6), and power (n ¼ 1) (see DB checklist). The number of agreements
in each category was entered in PASW (Version 18) (Predictive
Analytics Software, Armonk, NY) with total count weighted. The
Cohen’s Kappa (K) in 5 categories indicated that the agreement of
quality assessment was moderate in power (K ¼ 0.556; P ¼ .002),
substantial in reporting (K ¼ 0.796; P ¼ .000) and internal validity
(K ¼ 0.797; P ¼ .000), and outstanding in external validity (K ¼
0.850; P ¼ .000) and selection bias (K ¼ 0.830; P ¼ .000). The overall
Cohen’s Kappa was 0.782 (P ¼.000), demonstrating the agreement
between 2 raters was substantial. In addition to the DB checklist,
we also assessed the evidence-based, minimum set of recommen-
dations for reporting RCTs per Consolidated Standards of Reporting
Trials (CONSORT). Six studies adhered to CONSORT guidelines by
reporting intention to treat.10,14,27e30 Only one study reported
allocation concealment.17 After the completion of quality assess-
ment, each individual study’s variables were selected and entered
in the Comprehensive Meta-Analysis software (CMA, Englewood,
NJ) by the first author (M.C.), with said descriptors of each study
assigning it as either a high- or low-quality performance (Figure 2).
Statistical Methods
We used a random-effects model to determine the variance
associated with each of the sources of variation influencing the fall-
intervention programs. The random-effects model is more plau-
sible than the fixed effect model and has the benefit of reducing the
heterogeneity of sources.31 The outcome measure used for the
combined effect size was a risk ratio (RR) to estimate the overall
combined fall-reduction rate. The estimated RR of falls between
groups was computed, along with its variance and a standard error
of the RR, to yield a summary effect as the best approach consistent
with the primary outcome measures used in these studies. Statis-
tical heterogeneity was quantified using I2 and Q statistics. I2 is the
188.e15 M. Choi, M. Hector / JAMDA 13 (2012) 188.e13e188.e21

Titles and abstracts were a discernible difference between studies conducted where the
reviewed via electronic search
(n = 17,325)
intervention was accomplished and results were then monitored
Titles and abstracts unlikely to be relevant were (Model I) versus studies in which the intervention continued
excluded
(n = 15,419)
throughout the measured period (Model II). Sensitivity analyses
Studies that did not identify RCT were excluded were performed to assess the effect of excluding the trials for which
(n = 1,619)
studies showed high standardized random residual errors. Stan-
Potentially relevant studies
were reviewed dardized residual errors were calculated based on observed value
(n = 287) minus predicted value divided by the square root of the residual
Studies that included children falls were excluded
(n = 27) mean square. Publication bias was assessed using a funnel plot
Studies duplicated were excluded analysis.32 Mean effect was computed for this bias. Meta-regression
(n = 33)
Potentially relevant studies using mixed effects regression with unrestricted maximum likeli-
were reviewed hood was conducted to assess the degree of dispersion remaining
(n = 227)
Studies that did not report fall intervention or after sampling error variance.31 Proportion of total variance
effectiveness of the program were excluded explained by the model was calculated using T2total and T2unexplained.
(n = 151)
Studies that identified unclear intervention period or Comprehensive Meta-Analysis software (CMA, Version 2.2) was
did not meet the intervention period, and quasi used to synthesize the effectiveness of fall-intervention programs
experimental design studies were excluded
(n = 40) provided by the individual studies. Sensitivity analysis, a funnel
Studies that did not include both men and women plot analysis, and meta-regression were also performed using the
were excluded
(n = 15) CMA.

Potentially relevant studies

were reviewed
(n = 21)
Results
Studies completed with unreliable follow-up
were excluded
(n = 4) Study Characteristics
Final studies included in
the meta-analysis
(n = 17) The total sample size in the meta-analysis of the 17 studies was
5501 participants (281 for control and 2720 for intervention). We
Fig. 1. Flow chart of study selection.
used a sample size for each study pertinent to our meta-analysis
rather than total sample size reported in the study. For example,
Campbell et al13 reported a total sample size of 391 but we used
ratio of true heterogeneity to total variation in observed effects, a total sample size of 293 for the subgroup (home safety alone
reflecting the proportion of observed dispersion that is attributable group vs no home safety program) on the number of falls of 1 or
to heterogeneity. The Q statistic and its P value served as a test of more that met our criteria. Another example is Lord et al’s study33
significance.31 Subgroup analysis was performed by type (single vs in which the total sample size reported was 597 for control,
multifactorial intervention), setting (community vs nursing home), minimal, and extensive intervention groups. We used a total
and intervention intensity (Model I vs Model II). Models I and II sample size subset of 403 to compare the control and extensive
were our own construct in an attempt to discern if there was intervention groups. The samples in the meta-analysis had a mean

Study name Type Intervention Model Setting Statistics for each study Risk ratio and 95% CI
Risk Lower Upper
ratio limit limit Z-Value p-Value
Berggren Multi 1.000 Com 0.761 0.573 1.009 -1.896 0.058
Broe Single 2.000 NH 0.494 0.202 1.206 -1.548 0.122
Campbell Single 2.000 Com 1.000 0.778 1.285 -0.003 0.998
Clemson Multi 1.000 Com 0.828 0.614 1.118 -1.232 0.218
Davison Multi 2.000 Com 0.936 0.883 0.992 -2.237 0.025
Day Multi 1.000 Com 0.667 0.494 0.900 -2.647 0.008
Elley Multi 2.000 Com 1.096 0.932 1.288 1.106 0.269
Hendriks Multi 1.000 Com 1.017 0.671 1.542 0.080 0.936
Lightbody Multi 1.000 Com 0.853 0.792 0.919 -4.171 0.000
Lord Multi 2.000 Com 1.028 0.830 1.274 0.255 0.799
Mahoney Multi 1.000 Com 0.854 0.690 1.058 -1.440 0.150
Neyens Multi 2.000 NH 0.431 0.334 0.555 -6.500 0.000
Rapp Multi 2.000 NH 0.461 0.413 0.515 -13.675 0.000
Spice Multi 1.000 Com 0.899 0.811 0.998 -2.004 0.045
van Hasstregt Multi 2.000 Com 1.118 0.728 1.718 0.509 0.611
Vind Multi 1.000 Com 1.089 0.906 1.309 0.911 0.363
Whitehead Multi 2.000 Com 2.063 1.255 3.389 2.856 0.004
0.855 0.742 0.985 -2.168 0.030
0.1 0.2 0.5 1 2 5 10

RR = .855, z = -2.168, p = .030

2
Q =196.204, df = 16, p= .000, I = 91.845

Fig. 2. A total of 17 studies of random effects model.

 M. Choi, M. Hector / JAMDA 13 (2012) 188.e13e188.e21
age of 79.2. One study29 reported the mean age of the participants
was older than 70 and we treated this as a mean age of 70.
Type of Intervention
Single or multifactorial intervention in each study indicated
actual program management. In a study13 reported to be a multi-
faceted intervention, we used only a single component (eg, home
safety) of the multifaceted intervention to meet our inclusion
criteria and treated it as a single intervention. Each individual study
might include any of several types of intervention strategies that
were conducted with the goal of fall prevention. Two studies used
single intervention programs.13,15 Multifactorial intervention
programs included a comprehensive medical exam, occupational
therapy assessment, activities of daily living, home environmental
and behavioral assessment, cognition assessment, gait stability,
medication review, staff training, and education for residents.
Multifactorial team approaches included more than one staff
member, including physicians, nurses, physical therapists, occu-
pational therapists, social workers, and/or other trained health care
professionals. For example, individualized intervention programs
used in one study focused on counseling sessions after initial
assessment followed by individualized exercises aimed at
improving strength, coordination, and balance by trained accredi-
ted fitness instructors twice weekly was treated as a multifactorial
intervention.33 A total of 15 studies used multifactorial interven-
tions using multidisciplinary team approaches.
Setting of Intervention
One study’s intervention was applied in the hospital before
patient discharge but evaluated over 1 year of follow-up in the
community. We treated this study as a community intervention.17
One study sample recruited from community-dwelling individ-
uals who attended the emergency department of the university
hospital was treated as a community intervention because inter-
vention was home based.28 Three studies were conducted in
nursing homes.10,15,34 A total of 14 studies were conducted target-
ing community-dwelling residents. One study was a nurse-led
intervention,19 and 5 studies used home-based intervention
programs.18,19,28,35,36
Intensity of Intervention
Our initial intention was to determine the effectiveness of a fall-
prevention program between intervention periods shorter than
6 months versus longer than 12 months. Twelve studies used
shorter than 6-month interventions, and 14 studies used 12 months
or more of follow-up interventions. There were no other inter-
vention periods reported, such as 8 months or 9 months. There was
a great deal of variability in terms used (intervention period/
duration, data collection period, follow-up period/duration) across
the studies. For example, Elley et al18 reported that the exercise
intervention was delivered for 1 year during home visits at weeks 1,
2, 4, and 8, and after 6 months. However, no interventions were
reported after 6 months. We treated this study intervention period
as 6 months. Another study10 purporting a 12-month trial in
nursing homes reported that the intervention was 12 months with
12 months of follow-up, which counts as a total of 24 months.
Because the entire trial was 12 months, the intervention and
follow-up occurred at the same time. Furthermore, specific inter-
vention periods were not always reported. As an inference from this
fact, we thought it optimal to dichotomize interventions to Model I
or II to determine the effectiveness of fall-prevention programs.
We, therefore, divided all studies into Model I or II depending on
188.e16
the continuum of intervention (as a proxy for intervention inten-
sity). If an intervention was made in the beginning of a study and
not throughout the study, it was treated as a Model I. If the inter-
vention was continued throughout the study, it was treated as
a Model II. For instance, we treated studies reported as 2-hour
sessions for 7 weeks,14 2 visits at 4 months and 12 months,17
4-week intervention,19 2 home visits,37 intervention approxi-
mately continued 2.5 months,28 weekly exercise for 15 weeks,29
and 1-stop multidisciplinary clinic30 as Model I studies. The
follow-up period after intervention was also tabulated but not
included for analysis (Table 1). Most studies reported detailed
follow-up intervention periods and also used terms “intervention”
and “follow-up” interchangeably.
Pooled Results
The combined effect size based on RR and 95% confidence
intervals (CI) were calculated for each individual study and for the
overall 17 studies. The combined effect size of RR for the 17 studies
was 0.855 (z ¼ e2.168; P ¼.030; 95% CI ¼ 0.742e0.985). The overall
results indicate that the interventions across the studies were
effective by demonstrating a 14% reduction in falls (Figure 2),
implying that 1 fall of 7 is possibly preventable with an
intervention.
Subgroup Analyses
We found, however, that there was clear evidence of heteroge-
neity across the findings of the pooled results (Q ¼ 196.204; df ¼
16; P ¼ .000; I2 ¼ 91.845). Therefore, we attempted to analyze the
reasons for the variance by conducting group comparisons across
the studies. We conducted 3 different group analyses stratified by
(1) type of intervention: 2 studies for single interventions versus 15
studies for multifactorial interventions; (2) type of setting: 3
studies in nursing homes and 14 studies in the community; and (3)
intensity of intervention: 8 studies for Model I and 9 studies for
Model II. The subgroup analysis of all 17 studies demonstrated as
follows: a significant 14% fall reduction in multifactorial interven-
tions (RR ¼ 0.856; z ¼ e2.039; 95% CI ¼ 0.737e0.994; P ¼ .041) with
no study variations between the groups (Q ¼ 0.002; df ¼ 1; P ¼
.961); an inconceivable 54% fall reduction in nursing homes (RR ¼
0.453; z ¼ e9.366; 95% CI ¼ 0.384e0.535; P ¼ .000) with study
variations between the groups (Q ¼ 62.788; df ¼ 1; P ¼ .000); and
no significant fall reduction was identified in either Model I or II.
Sensitivity Analyses
A sensitivity analysis is crucial in determining how robust the
findings were.31 We were interested in how the findings would
shift if we determined results deleting high standardized random
residuals. Four of the 17 studies showed large standard residuals
ranging from e2.42 to 2.46. These studies were also identified as
outliers whose effects were substantially different from others.
According to Borenstein et al,31 sensitivity analysis can be used “to
investigate whether the conclusions would differ across a range of
plausible imputed values.” We removed these 4 studies one at
a time from the analysis to determine the change of overall effect
size until the homogeneity across the studies was determined. The
combined effect size of RR of 13 studies after excluding 4
studies10,18,34,38 was 0.906 (z ¼ e3.179; P ¼ .001), with 95% CI of
0.853e0.963, demonstrating there was an overall significant fall-
reduction rate (9%) across the studies (Figure 3). These remaining
13 studies were found to be homogeneous (Q ¼ 18.582; df ¼ 12; P ¼
.099; I2 ¼ 35.423), demonstrating that there was no variation across
the studies.
 188.e17
Table 1
Studies Included in Meta-Analysis

First Author Year Control Intervention Intervention Intervention Setting Mean Intervention Programs
n n Period/Follow- Model Age
up, mo
1. Berggren17 2008 97 102 1/12 I C 81 Two visits at 4 months and 12 months focusing on comprehensive geriatric assessment of in-hospital
patients after hip fracture, calcium 1000 mg with vitamin D 800 IU, rehab program, home training programs (M)
2. Broe15 2007 25 23 5/5 II NH 89 Simple vitamin D supplement with different doses in 4 groups for 5 months (S)
3. Campbell13 2005 196 97 1/12 II C 84 Two-day home safety alone program focusing on muscle strengthening and balance retraining for elderly with
severe visual impairments (S)
4. Clemson14,* 2004 153 157 2/14 I C 76 Two-hour session for 7 weeks focusing on lower limb balance and strengthening exercise, improving home and
community environmental and behavioral safety, making adaptations to low vision, and medication review (M)
27,*
5. Davison 2005 154 159 12/12 II C 77 Follow-up every 4 weeks over 12 months with multifactorial intervention focusing on comprehensive cardio,

M. Choi, M. Hector / JAMDA 13 (2012) 188.e13e188.e21

cognitive, medication, and vision assessment with balance and strengthening exercise provided by a university
hospital (M)
6. Day29,* 2002 135 137 4/18 I C >70 Weekly exercise class of 1 hour for 15 weeks using group-based strength and balance exercise, home hazard
management, and vision improvement program for outpatients (M)
7. Elley18 2008 157 155 6/12 II C 81 Five home-visit-exercise at weeks 1, 2, 4, and 8, and after 6 months focusing on comprehensive community
intervention program including health, home hazards, bone health assessment, and exercise program (M)
28,*
8. Hendriks 2008 134 124 3/12 I C 75 Intervention approximately 2.5 months (with maximum 3.5 months) focusing on comprehensive outpatient
geriatric assessment, structured functional and environmental assessment by OT, FAI and OT checklist provided
by ED of hospital, and recommendations reinforced for behavioral change, functional needs and safety (M)
9. Lightbody19 2002 177 171 1/6 I C 75 Four-week interventions including home visit and assessment focusing on medications, ECG, BP, cognition,
visual acuity, vestibular dysfunction, balance, mobility, feet and foot wear, social service referrals, modification
of risk factors (M)
10. Lord33 2005 201 202 12/12 II C 80 Individualized outpatient program focusing on strengthening exercise, maximizing vision and sensation provided
by hospital fall clinic (M)
37
11. Mahoney 2007 175 174 1/12 I C 80 Two home visits by nurse and PT focusing on multifactorial intervention program that links participants to existing
medical care and service network including balance exercise and 11 monthly phone calls (M)
12. Neyens10,* 2009 269 249 12/12 II NH 83 Multidisciplinary fall-prevention programs focusing on fall risks, medication intake, mobility, and the use of assistive
and protective aids in demented patients in psychogeriatric nursing homes (M)
13. Rapp34 2008 360 365 12/6 II NH 86 Comprehensive environmental assessment, group exercise program twice/week using 6-month follow-up, staff
training, organizational approach in nursing home patients (M)
30,*
14. Spice 2009 159 210 1/12 I C 82 One-stop multidisciplinary clinic with referrals for investigations, interventions including home check and
follow-up (M)
15. van 2000 123 129 12/18 II C 77 Five home visits over a period of a year focusing on a structural protocol including medical, environmental, and
Haastregt36 behavioral factors causing falls (M)
16. Vind35 2009 196 196 1/12 I C 74 Multifactorial individualized intervention program during 2 initial visits at geriatric outpatient fall clinic focusing
on MMSE, GDS, vision, vestibular deficiency, ECG, Dual-energy X-ray assessment, and orthostatic BP measure,
patients were telephoned monthly for collection of fall data (M)
17. Whitehead38 2003 70 70 6/6 II C 78 Interdisciplinary outpatient exercise program provided during 6 months of follow-up by community- and
hospital-based services focusing on falls and balance, home assessment, medication reduction program (M)

n, sample size; C, community; NH, nursing home; H, hospital; OT, occupational therapy; FAI, Frenchay Activity Index; ED, emergency department; ECG, electrocardiogram; BP, blood pressure; M, multifactorial intervention;
S, single intervention; MMSE, Mini-Mental State Examination; GDS, Geriatric Depression Scale.
*
Indicates intention-to-treat analysis reported in each study.
 M. Choi, M. Hector / JAMDA 13 (2012) 188.e13e188.e21 188.e18

Study name Type Intervention Model Setting Statistics for each study Risk ratio and 95% CI
Risk Lower Upper
ratio limit limit Z-Value p-Value
Berggren Multi 1.000 Com 0.761 0.573 1.009 -1.896 0.058
Broe Single 2.000 NH 0.494 0.202 1.206 -1.548 0.122
Campbell Single 2.000 Com 1.000 0.778 1.285 -0.003 0.998
Clemson Multi 1.000 Com 0.828 0.614 1.118 -1.232 0.218
Davison Multi 2.000 Com 0.936 0.883 0.992 -2.237 0.025
Day Multi 1.000 Com 0.667 0.494 0.900 -2.647 0.008
Hendriks Multi 1.000 Com 1.017 0.671 1.542 0.080 0.936
Lightbody Multi 1.000 Com 0.853 0.792 0.919 -4.171 0.000
Lord Multi 2.000 Com 1.028 0.830 1.274 0.255 0.799
Mahoney Multi 1.000 Com 0.854 0.690 1.058 -1.440 0.150
Spice Multi 1.000 Com 0.899 0.811 0.998 -2.004 0.045
Vind Multi 1.000 Com 1.089 0.906 1.309 0.911 0.363
van Hasstregt Multi 2.000 Com 1.118 0.728 1.718 0.509 0.611
0.906 0.853 0.963 -3.179 0.001
0.1 0.2 0.5 1 2 5 10

RR = .906, z =-3.179, p = .001

Q = 18.582, df = 12,p= .099, I 2 = 35.423

Fig. 3. Sensitivity analysis of random effects model using a total of 13 studies.

We also performed subgroup analyses using the 13 homoge- Publication Bias

neous studies to determine any differences in fall reduction by type,
setting, and intensity of intervention. The combined effect size of
RR for multifactorial interventions was 0.904 (z ¼ e3.036; P ¼ .002;
95% CI ¼ 0.847e0.965), demonstrating that a significant 10% fall
reduction was found with no study variation between the groups
(Q ¼ 0.077, df ¼ 1, P ¼ .782). The subgroup analysis by setting
indicated that there was a significant 9% fall reduction in the
community (RR ¼ 0.909; z ¼ e3.179; P ¼ .001; 95% CI ¼
0.856e0.964) with no study variations between the groups (Q ¼
1.757, df ¼ 1, P ¼ .185). The subgroup analysis by intervention
intensity demonstrated that a significant 12% fall reduction was
found in Model I (RR ¼ 0.876, 95% CI ¼ 0.815e0.943) with no study
variations between the groups (Q ¼ 2.168, df ¼ 1, P ¼ .141).
We used a funnel plot to determine the relationship between
study sample size and effect size. The funnel plot was plotted
with effect size using log risk ratio on the x-axis and the standard
error on the y-axis. The 17 open circles indicate observed point
estimates. Four observed point estimates are shown out of the
funnel chart demonstrating substantive publication bias. Because
of the relatively small number of studies (n ¼ 17), we expected
that these studies have more sampling error variation in effect
and tend to be spread asymmetrically across a broad range of
values (Figure 4). The result indicated that there was evidence of
bias in favor of publication of a positive result in the meta-
analysis.

Funnel Plot of Standard Error by Log risk ratio

0.0

0.1
Standard Error

0.2

0.3

0.4

0.5

-2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0

Log risk ratio

Fig. 4. Funnel plot of 17 studies observed and imputed.

188.e19
0.80
0.62
0.44
0.26
Log risk ratio
0.08
-0.10
-0.28
-0.46
-0.64
-0.82
-1.00
-11.20 29.84
Fig. 5. Meta-regression of 17 studies using sample size on log risk ratio.
Meta-regression
We next conducted meta-regression to determine the impact of
covariate (sample size) for statistical significance because it is
important to quantify the magnitude of the relationship with effect
size. For this reason, we used R2 as an index based on the percent
reduction in true variance. Because studies included in our meta-
regression are performed by different researchers using different
methodologies, it is more reasonable to use mixed effects regres-
sion (random effects model: unrestricted maximum likelihood).
Under the regression model, it is assumed that all studies that have
the same value on the sample size share a common effect size. The
amount of total variance was more than we expected based on
within-study error (Q ¼ 196.204; df ¼16; P ¼ .000). The relationship
between log risk ratio and sample size was significant (Q model ¼
131.953; df ¼1; P¼ .000); some of the between-study variance
remains unexplained (Q resid ¼ 64.250; df ¼15; P ¼ .000) as depicted
in Figure 5. Proportion of total variance explained by the model is
calculated by R2 ¼ 1 e (T2unexplained / T2total ). T2total was 0.069 and
T2unexplained was 0.021. Overall, the model fit explained 68.6% of the
relevant variance, demonstrating a moderate model fit (Figure 5).
Discussion
Our final selected 17 studies using RCTs focusing on fall-
prevention programs were analyzed using RR to determine the
effectiveness of prevention programs in older adults. Overall,
a significant 14% fall reduction was found on the number of falls and
fall rate during the follow-up along with an even more significant
fall reduction in multifactorial intervention studies (n ¼ 15, 14%)
and nursing homes (n ¼ 3, 55%). The significant fall-prevention
effect in multifactorial intervention program in our analysis was
consistent with the results of several prior studies7,9,21,39,40 and
inconsistent with 2 earlier meta-analysis efforts20,21 that reported
that a single intervention specifically focusing on exercise alone
was more effective in reducing falls. Previously, a significant fall-
prevention intervention effect in nursing homes was found in
earlier studies,5,21 which was consistent with our study result in
nursing home fall reduction. The intensity of intervention did not
demonstrate any significant fall reduction rate either in Model I
or II.
Under the random effects model, the confidence intervals pre-
sented for the effect size in 17 studies were wide and, thus,
sampling error and between-studies variance cannot be avoided;
M. Choi, M. Hector / JAMDA 13 (2012) 188.e13e188.e21
Regression of Sample size on Log risk ratio
70.88 111.92 152.96 194.00 235.04 276.08 317.12 358.16 399.20
Sample size
this resulted in heterogeneity across the overall studies. Because of
the heterogeneity, as well as limited sample size in the studies in
nursing homes (n ¼ 3), with such significant variation between the
groups, we could not conclude that (1) the overall fall-reduction
results were valid, or that (2) the multifactorial intervention
program or (3) the intervention in the nursing home setting was
effective in reducing fall rates in the elderly.
Therefore, further sensitivity analysis was performed to examine
how the results would shift if we eliminated studies with high
standardized random residuals. After elimination of 4 studies with
high standardized random residuals,10,18,34,38 the overall studies
showed that the fall-intervention programs were significantly
effective in multifactorial interventions (10%, n ¼ 11), as were studies
done in community settings (9%, n ¼ 12) and when the intervention
intensity did not continue through the entire study period (12%, n ¼
8). The result also demonstrated that there were no variations across
the remaining 13 studies and between the groups.
Previous studies have noted that there have been other
meta-analysis studies reporting a lack of effectiveness in
fall-intervention programs targeting older adults.7,41,42 In 2002,
Hill-Westmoreland et al’s meta-analysis41 reported an inconse-
quential (4%) decrease in the rate of falls for individuals who were
in the treatment groups and concluded that more studies are
needed. In 2004, Weatherall’s meta-analysis conclusion42 was that
evidence for exercise as a sole intervention was neither for nor
against falls prevention. Other studies report that not only are fall-
intervention programs successful, but that the most effective fall-
prevention programs were based on multifactorial interventions,
including systemic fall-risk assessments, ongoing staff education,
safety checks, environmental modifications, and targeted exercise
programs.7,9,21,39,40 Robertson et al’s meta-analysis39 suggested that
a customized and supervised home exercise program is of benefit
for the community elderly; Campbell and Robertson’s meta-
regression analysis21 concluded that multifactorial fall-prevention
interventions are effective for individuals, but for community
programs for populations at risk, targeted single interventions are
as effective, and might be more acceptable and cost effective. Pet-
ridou et al’s 2009 meta-analysis20 concluded that exercise-alone
single-intervention studies were 5 times more effective compared
with studies focused on multifactorial prevention programs. Indi-
vidualized and targeted exercise programs and/or other prevention
programs combined with exercise focusing on balance, progressive
resistance training, and lower-body strength were also reported as
being effective in selected meta-analyses.11,12,22
 M. Choi, M. Hector / JAMDA 13 (2012) 188.e13e188.e21
There is little disagreement that RCTs are the most rigorously
designed scientific experimental studies in testing such questions
as these 17 trials on fall prevention attempt to address. Yet, in
fall-intervention RCT studies, blinding the control group to such
prevalent ambient sources of information about evolving
practices and assistive equipment is nearly impossible. None
of the studies reported blinding. Unlike blinding, studies
with allocation concealment reported to reduce selection bias
and protect the randomization were few. It is for these and
undoubtedly many more known or unknown reasons that it
becomes so difficultdbordering on the heroicdto do essential fall-
intervention research studies. Therefore, it is difficult to examine the
literature in aggregate and derive an intervention, or set of inter-
ventions, that truly seems to lend itself to recommendations that
providers can efficiently incorporate into their practice.
Limitation
We found significant differences among the studies included
in our meta-analysis. Owing to their small number, one potential
problem might be estimating the between-study variance, which
might in turn have an influence on common effect sizes. Several
potential limitations of this meta-analysis need to be addressed.
The limitation of a search based on electronic databases might
lead to an overestimation or underestimation of effect sizes. We
also consider that one significant limitation of the existing
evidence is the lack of sample size because of our rigorous
criteria. The fact that several of these 17 studies concluded on
their own that their intervention made no significant difference
in the outcome and is not to be recommended was reassuring,
however, of our analysis. The variety of outcome measurement
methods used to determine and report patient fall rates in our
selected studies made it difficult to compare overall fall rates
between institutions and patient care. Last, the quality assess-
ment by a group of trained reviewers with a confirmation by
a second group of reviewers before an article’s publication would
also be helpful in increasing the precision of the consistency of
study findings in the future.
Conclusions
The specific aim of the study was to examine the effectiveness of
fall-prevention programs conducted for older adults using RCTs
published in the most recent 10 years. Based on meta-sensitivity
analyses, we conclude that interventions to prevent falls in older
adults were modestly effective, in multifactorial intervention with
a 10% reduction in fall rates, 9% in community settings, and 12% on
Model I intervention types (initial intervention and subsequent
follow-up).
We found a few observations that might be valuable for future
research. Studies need to be carefully designed to examine their
results over an extended period of time (at least 6 months) if they
are to be meaningful. Excluding observational studies with signif-
icant variations of outcome measures would help to reduce
confounders. Eliminating quasi-experimental studies would help to
yield an unbiased estimate of the effect size. Including only studies
of RCTs may reduce publication bias. We certainly would encourage
the inclusion of more randomized controlled studies to reduce
sampling error variation, even if the results turned out not to be
“positive.”
Defining what is the most efficient and effective intervention in
terms of fall prevention for this highly at-risk population remains
a formidable task. Despite the best of management, all falls are not
preventable. One must presume that with the preponderance of
medical, financial, and legal incentives to prevent falls, the “low-
188.e20
hanging fruit” of falls preventiondbed, chair, and personal alarms;
hip protection devices; lowered beds; mats on the floor; and
meticulous training of attendants to pay attention to all of these to
the extent that they are of benefitdhave already been incorporated
into a necessarily practical prevention program.
Implications for Clinical Practice
Implementing intervention programs preventing falls for older
adults seems to be plausible and highly desirable, but the overall
effectiveness of intervention programs are not supported strongly
by significant statistical results. The following are recommenda-
tions for health care providers to attempt to reduce fall rates in
their clinical practice: (1) identify an individual’s risk factors for
falls; (2) determine predisposing and precipitating factors if the
patient has a history of falls, and intervene accordingly; (3)
provide intervention programs and management focusing on
lower-extremity balance and strengthening; (4) consider
psychological factors such as fear of falling and self-imposed
restriction of activity;10 and (5) classify injuries when they do
occur based on the International Classification of Diseases (10th
revision, classification system).40
References
1. National Center for Injury Prevention and Control. CDC injury fact book. 2006.
Available at: http://www.cdc.gov/Injury/Publications/FactBook. Accessed
October 10, 2010.
2. Centers for Disease Control and Prevention. Hip fractures among older adults.
2009. Available at: http://www.cdc.gov/homeandrecreationalsafety/falls/index.
html. Accessed December 7, 2010.
3. Centers for Disease Control and Prevention. Cost of falls among older adults.
2009. Available at: http://www.cdc.gov/HomeandRecreationalSafety/Falls/
fallcost.html. Accessed November 2, 2010.
4. Stevens JA, Thomas K, Teh L, Greenspan AI. Unintentional fall injuries associ-
ated with walkers and canes in older adults treated in U.S. emergency
departments. J Am Geriatr Soc 2009;57:1464e1469.
5. Gates S, Fisher JD, Cooke MW, et al. Multifactorial assessment and targeted
intervention for preventing falls and injuries among older people in commu-
nity and emergency care settings: Systematic review and meta-analysis. BMJ
2008;336:130e133.
6. Gillespie LD, Gillespie WJ, Robertson MC, et al. Interventions for preventing
falls in elderly people. Cochrane Database Syst Rev 2003;(4):CD000340.
7. Chang JT, Morton SC, Rubenstein LZ, et al. Interventions for the prevention of
falls in older adults: Systematic review and meta-analysis of randomised
clinical trials. BMJ 2004;328:680e683.
8. Oliver D, Connelly JB, Victor CR, et al. Strategies to prevent falls and fractures in
hospitals and care homes and effect of cognitive impairment: Systematic
review and meta-analyses. BMJ 2007;334:82e87.
9. Rubenstein LZ. Falls in older people: Epidemiology, risk factors and strategies
for prevention. Age Ageing 2006;35:ii37eii41.
10. Neyens JCL, Dijcks BP, Twisk J, et al. A multifactorial intervention for the
prevention of falls in psychogeriatric nursing home patients, a randomized
controlled trial (RCT). Age Ageing 2009;38:194e199.
11. Gillespie L. Preventing falls in elderly people. BMJ 2004;328:653e654.
12. Dukyoo J, Juhee L, Lee SM. A meta-analysis of fear of falling treatment programs
for the elderly. West J Nurs Res 2009;31:6e16.
13. Campbell AJ, Robertson MC, La Grow SJ, et al. Randomised controlled trial of
prevention of falls in people aged > or ¼75 with severe visual impairment: The
VIP trial. BMJ 2005;331:817e820.
14. Clemson L, Cumming RG, Kendig H, et al. The effectiveness of a community-
based program for reducing the incidence of falls in the elderly: A random-
ized trial. J Am Geriatr Soc 2004;52:1487e1494.
15. Broe KE, Chen TC, Weinberg J, et al. A higher dose of vitamin D reduces the risk
of falls in nursing home residents: A randomized, multiple-dose study. J Am
Geriatr Soc 2007;55:234e239.
16. Steinberg M, Cartwright C, Peel N, et al. A sustainable programme to prevent
falls and near falls in community dwelling older people: Results of a rando-
mised trial. J Epidemiol Community Health 2000;54:227e232.
17. Berggren M, Stenvall M, Olofsson B, et al. Evaluation of a fall-prevention
program in older people after femoral neck fracture: A one-year follow-up.
Osteoporos Int 2008;19:801e809.
18. Elley CR, Robertson MC, Garrett S, et al. Effectiveness of a falls-and-fracture
nurse coordinator to reduce falls: A randomized, controlled trial of at-risk
older adults. J Am Geriatr Soc 2008;56:1383e1389.
188.e21 M. Choi, M. Hector / JAMDA 13 (2012) 188.e13e188.e21
19. Lightbody E, Watkins C, Leathley M, et al. Evaluation of a nurse-led falls
prevention programme versus usual care: A randomized controlled trial. Age
Ageing 2002;31:203e210.
20. Petridou ET, Manti EG, Ntinapogias AG, et al. What works better for
community-dwelling older people at risk to fall? A meta-analysis of multi-
factorial versus physical exercise-alone interventions. J Aging Health 2009;
21:713e729.
21. Campbell AJ, Robertson MC. Rethinking individual and community fall
prevention strategies: A meta-regression comparing single and multifactorial
interventions. Age Ageing 2007;36:656e662.
22. Sherrington C, Whitney JC, Lord SR, et al. Effective exercise for the prevention
of falls: A systematic review and meta-analysis. J Am Geriatr Soc 2008;56:
2234e2243.
23. Coussement J, De Paepe L, Schwendimann R, et al. Interventions for preventing
falls in acute- and chronic-care hospitals: A systematic review and meta-
analysis. J Am Geriatr Soc 2008;56:29e36.
24. Hendriks MR, Bleijlevens MH, van Haastregt JC, et al. Lack of effective-
ness of a multidisciplinary fall-prevention program in elderly people
at risk: A randomized, controlled trial. J Am Geriatr Soc 2008;56:
1390e1397.
25. Downs SH, Black N. The feasibility of creating a checklist for the assessment of the
methodological quality both of randomised and non-randomised studies of
health care interventions. J Epidemiol Community Health 1998;52:377e384.
26. Spitzer RL, Cohen JA, Fliss JL. Quantification of agreement in psychiatric diag-
nosis: A new approach. Arch Gen Psychiatry 1968;17:83e87.
27. Davison J, Bond J, Dawson P, et al. Patients with recurrent falls attending
accident & emergency benefit from multifactorial interventionda randomised
controlled trial. Age Ageing 2005;34:162e168.
28. Hendriks MR, Bleijlevens MH, van Haastregt JC, et al. Lack of effectiveness of
a multidisciplinary fall-prevention program in elderly people at risk: A
randomized, controlled trial. J Am Geriatr Soc 2008;56:1390e1397.
29. Day L, Fildes B, Gordon I, et al. Randomised factorial trial of falls prevention
among older people living in their own homes. BMJ 2002;325:128e131.
30. Spice CL, Morotti W, George S, et al. The Winchester falls project: A randomized
controlled trial of secondary prevention of falls in older people. Age Ageing
2009;38:33e40.
31. Borenstein M, Hedges LV, Higgins JPT, Rothstein HR. Introduction to Meta-
Analysis. Chichester, United Kingdom: John Wiley & Sons, Ltd; 2009.
32. Light RJ, Singer JD, Willett JB. The visual presentation and interpretation of
meta-analyses. In: Cooper IM, Hedges LV, editors. The Handbook of Research
Synthesis. New York, NY: Russell Sage Foundation; 1994.
33. Lord SR, Tiedemann A, Chapman GK, et al. The effect of an individualized fall
prevention program on fall risk and falls in older people: A randomized,
controlled trial. J Am Geriatr Soc 2005;53:1296e1304.
34. Rapp K, Lamb SE, Buchele G, et al. Prevention of falls in nursing homes:
Subgroup analyses of a randomized fall prevention trial. J Am Geriatr Soc 2008;
56:1092e1097.
35. Vind AB, Andersen HE, Pedersen KD, et al. An outpatient multifactorial falls
prevention intervention does not reduce falls in high-risk elderly Danes. J Am
Geriatr Soc 2009;57:971e977.
36. van Haastregt JC, Diederiks JPM, van Rossum E, et al. Effects of a programme of
multifactorial home visits on falls and mobility impairments in elderly people
at risk: Randomised controlled trial. BMJ 2000;321:994e998.
37. Mahoney JE, Shea TA, Przybelski R, et al. Kenosha county falls prevention study: A
randomized, controlled trial of an intermediate-intensity, community-based
multifactorial falls intervention. J Am Geriatr Soc 2007;55:489e498.
38. Whitehead C, Wundke R, Crotty M, et al. Evidence-based clinical practice in
falls prevention: A randomised controlled trial of a falls prevention service.
Aust Health Rev 2003;26:88e97.
39. Robertson MC, Campbell AJ, Gardner MM, et al. Preventing injuries in older
people by preventing falls: A meta-analysis of individual-level data. J Am
Geriatr Soc 2002;50:905e911.
40. Lamb SE, Jorstad-Stein EC, Hauer K, et al. Prevention of Falls Network Europe
and Outcomes Consensus Group. Development of a common outcome data set
for fall injury prevention trials: The prevention of falls network Europe
consensus. J Am Geriatr Soc 2005;53:1618e1622.
41. Hill-Westmoreland EE, Soeken K, Spellbring AM. Ameta-analysis of fall
prevention programs for the elderly: How effective are they? Nurs Res 2002;
51:1e8.
42. Weatherall M. Prevention of falls and fall-related fractures in communi-
tydwelling older adults: A meta-analysis of estimates of effectiveness based on
recent guidelines. Intern Med J 2004;34:102e108.