ORIGINAL INVESTIGATION

ONLINE FIRST
Patient Education to Prevent Falls Among
Older Hospital Inpatients
A Randomized Controlled Trial
Terry P. Haines, PhD; Anne-Marie Hill, MS; Keith D. Hill, PhD; Steven McPhail, BS; David Oliver, MD;
Sandra Brauer, PhD; Tammy Hoffmann, PhD; Christopher Beer, MBBS

Background: Falls are a common adverse event dur- als only, 8.61; and complete program, 7.63). However,
ing hospitalization of older adults, and few interven- there was a significant interaction between the interven-
tions have been shown to prevent them. tion and presence of cognitive impairment. Falls were less
frequent among cognitively intact patients in the com-
Methods: This study was a 3-group randomized trial to
plete program group (4.01 per 1000 patient-days) than
evaluate the efficacy of 2 forms of multimedia patient edu-
cation compared with usual care for the prevention of among cognitively intact patients in the materials-only
in-hospital falls. Older hospital patients (n = 1206) ad- group (8.18 per 1000 patient-days) (adjusted hazard ra-
mitted to a mixture of acute (orthopedic, respiratory, and tio, 0.51; 95% confidence interval, 0.28-0.93]) and con-
medical) and subacute (geriatric and neurorehabilita- trol group (8.72 per 1000 patient-days) (adjusted haz-
tion) hospital wards at 2 Australian hospitals were re- ard ratio, 0.43; 95% confidence interval, 0.24-0.78).
cruited between January 2008 and April 2009. The in-
Conclusion: Multimedia patient education with trained
terventions were a multimedia patient education program
based on the health-belief model combined with trained health professional follow-up reduced falls among pa-
health professional follow-up (complete program), multi- tients with intact cognitive function admitted to a range
media patient education materials alone (materials only), of hospital wards.
and usual care (control). Falls data were collected by Trial Registration: anzctr.org.au Identifier:
blinded research assistants by reviewing hospital inci-
ACTRN12608000015347
dent reports, hand searching medical records, and con-
ducting weekly patient interviews.
Arch Intern Med. 2011;171(6):516-524.
Results: Rates of falls per 1000 patient-days did not dif- Published online November 22, 2010.
fer significantly between groups (control, 9.27; materi- doi:10.1001/archinternmed.2010.444

F
ALLS ARE A LEADING PATIENT peared effective for preventing falls in
safety incident event in gen- hospitals, no recommendations could be
eral hospitals and are espe- made regarding effective components of
cially common in older pa- these multifactorial interventions. In ad-
tients.1 Approximately 30% of dition, compared with individual inter-
falls result in injury,2 the consequences of ventions, multifactorial falls programs may
which may cause increased length of stay (1) be more difficult and costly to imple-
or risk of institutionalization for the pa- ment, (2) create confusion for individual
tient,3 and legal complaint with subse- patients, and (3) reduce the effectiveness
quent litigation against the health service.4 of constituent components.13 Hence, there
is need to identify single intervention strat-
See also page 525 egies that prevent falls across a mixture of
hospital wards.
A promising intervention is the patient
Randomized trials of single interven- education program used as a part of the first
tions to prevent falls in hospitals have not targeted multifactorial program shown to
identified a statistically significant reduc- prevent falls in a randomized trial.2 The
tion in falls outcomes.5-8 Multifactorial in- education program involved providing
terventions have also been investigated written information coupled with 1-to-1
with mixed results.2,9-11 A recent Coch- follow-up with a research occupational
Author Affiliations are listed at rane review of these trials12 found that al- therapist. Surprisingly, exploratory analy-
the end of this article. though multifactorial interventions ap- ses revealed that the intervention was ef-

ARCH INTERN MED/ VOL 171 (NO. 6), MAR 28, 2011 WWW.ARCHINTERNMED.COM
516

©2011 American Medical Association. All rights reserved.

Corrected on March 28, 2011
Downloaded From: https://jamanetwork.com/ by Khalid Alomari on 07/08/2023
 fective (~50% reduction) in people with impaired cogni- Usual ward-based care varied between and within hospitals
tive function as well as in those with intact cognitive though it primarily consisted of falls risk screening using lo-
function, although contamination by other interventions cally developed instruments, use of risk alert items (arm bands),
included in that trial clouds these results.14 This finding led generic interventions (eg, nursing checklist to prompt activi-
ties such as a regular toileting program and regular visual ob-
authors to recommend further investigation of patient edu-
servation of patients), and additional 1-to-1 nursing for pa-
cation in isolation to determine if this intervention was ef- tients with acute agitation and/or confusion at extreme risk of
fective in isolation and equally effective for patients who falls. Physical restraint was not a front-line method for man-
have intact vs those with impaired cognitive function. The aging patients with agitation and/or confusion at either of the
present study addresses this recommendation by compar- participating sites. Multidisciplinary input (eg, medical, nurs-
ing 2 models of providing patient education to prevent in- ing, physiotherapy, occupational therapy) was routinely pro-
hospital falls vs usual care. vided on all wards, although therapists such as physiothera-
pists and occupational therapists provided more intensive input
(ie, daily 1-hour sessions) on subacute rehabilitation wards.
METHODS

MEASUREMENTS
DESIGN, PARTICIPANTS, AND SETTING
The primary outcome measure was participant falls. The defi-
A 3-group randomized controlled trial with recruiters, data col- nition of a fall used in this study was the World Health Orga-
lectors, and statistical analyst blind to group allocation, was un- nization definition: “an event which results in a person com-
dertaken.15 Potential participants were older adults admitted ing to rest inadvertently on the ground or floor or other lower
to acute (orthopedic and acute-respiratory medicine) and sub- level.”19 Prestudy training was provided to hospital staff on study
acute (geriatric assessment and rehabilitation) wards of the Prin- wards regarding classification of falls and procedures for re-
cess Alexandra Hospital, Brisbane, Australia, and the acute cording falls on incident reports using previously developed
(medical-surgical) and subacute (restorative–stroke rehabili- video materials.20 Falls data were collated from 3 sources dur-
tation) wards of Swan Districts Hospital, Perth, Australia. Pa- ing the trial: computerized incident reports, hand searching of
tients were excluded if (1) they were too ill to provide in- individual patient medical notes, and weekly patient inter-
formed consent, as determined by hospital staff, until discharge, views (or at patient discharge if earlier than 1 week), and falls
death, or transfer to a nonstudy ward; or (2) if they had pre- captured through any of these approaches were included. It was
viously participated in the trial. considered important to use multiple sources of data collec-
tion for the primary outcome owing to identified limitations
INTERVENTIONS of using single sources.21-23
Numerous participant demographic measures were taken
Two models of a patient education program were tested. The at the baseline assessment, including the Short Portable Men-
first (complete program) involved providing written and video- tal Status Questionnaire (SPMSQ)24 as a screen of cognitive func-
based materials and 1-to-1 follow-up with a health profes- tion where scores of 7 of 10 or below indicated impairment.
sional (physiotherapist) trained to provide this program at the This cut point corresponded to 23 of 30 or below on the Mini-
patient’s bedside. The content and progression of this educa- Mental State Examination (the cut point used in the previous
tion program was based on the health-belief model and in- subgroup analysis of the education program14) when 455 avail-
cluded presentation of epidemiologic falls data (frequency and able Mini-Mental State Examination scores were regressed
outcomes), causes of falls, self-reflection of individual risk, prob- against SPMSQ scores from this baseline demographic data set.
lem area identification, development of preventive strategies The Geriatric Depression Scale25 and the EQ-5D26 (formerly Eu-
and behaviors, goal setting, and goal review.16 Video materials roQol) health-related quality of life instruments were also ad-
were subjected to extensive testing and consumer feedback,17 ministered (Table 1).
and the overall program underwent incremental cost- Time spent by trained health professionals providing the com-
effectiveness analysis economic modeling to ensure feasibility plete program was recorded session by session. The trained health
of the delivery approach.18 Video materials were viewed by pa- professional at the Princess Alexandra Hospital site also re-
tients using a portable digital video disk player with a 9-inch corded the written behavior modification goals that were set by
screen and external head phones. Bedside curtains were drawn participants in the complete program and materials-only groups.
during the 1-to-1 follow-up to minimize contamination with
participants not allocated to this group. One-to-1 follow-up ses- PROCEDURE
sions were aimed to be completed during the first week of pa-
tient involvement in the study. The number of actual sessions Recruitment
provided was at the discretion of the research physiothera-
pists providing the follow-up.
Participant flow through this study is presented in Figure 1.
The second model (materials only) involved providing the
All patients admitted to subacute study wards were referred to
written and video-based materials without the trained health
researchers by clinical staff. Patients older than 60 years on acute
professional follow-up. Assistance was provided by the trained
wards who were expected to stay at least 3 more days were also
health professional to use the portable digital video disk player
referred. Those referred were approached for consent by re-
for viewing of the video materials.
searchers to participate as soon as practicable. Family mem-
Both interventions were provided in addition to usual ward-
bers were approached for consent where treating clinicians had
based care.
assessed the patient to have impaired cognitive function. Re-
cruitment occurred between January 2008 and April 2009, with
CONTROL the final participant being discharged in October 2009. Par-
ticipants recruited on one ward but later transferred to an-
A usual-care-only control group (control) received no specific other ward participating in this study (eg, transferred from acute
falls prevention education from the research team members. ward to rehabilitation ward) were observed until discharge to

ARCH INTERN MED/ VOL 171 (NO. 6), MAR 28, 2011 WWW.ARCHINTERNMED.COM
517

©2011 American Medical Association. All rights reserved.

Corrected on March 28, 2011
Downloaded From: https://jamanetwork.com/ by Khalid Alomari on 07/08/2023
 Table 1. Participant Demographics and Baseline Assessment Outcomes a

Characteristic Control Materials Only Complete Program

Total No. 381 424 401
Recruited on acute study wards 257 (67) 254 (60) 258 (67)
Recruited on subacute study wards 124 (33) 170 (40) 143 (33)
Transferred from acute study ward to subacute study ward during trial 15 (4) 11 (3) 11 (3)
Age, mean (SD), y 75.3 (10.1) 74.7 (11.7) 75.3 (11.0)
Male sex 178 (47) 201 (47) 185 (46)
Diagnosis group
Stroke 29 (8) 41 (10) 28 (7)
Orthopedic 136 (36) 160 (38) 151 (38)
Pulmonary 66 (17) 47 (11) 55 (14)
Other geriatric management b 46 (12) 46 (11) 41 (10)
All other diagnoses combined 104 (27) 130 (31) 126 (31)
English as a first language 348 (91) 376 (89) 359 (90)
Highest educational level attained
Primary school (up to age 11-12 y) 114 (30) 120 (28) 111 (28)
Year 10 163 (43) 189 (45) 171 (43)
Year 12 38 (10) 46 (11) 47 (12)
Tertiary 65 (17) 69 (16) 72 (18)
Premorbid living arrangements
Community alone 136 (36) 148 (35) 140 (35)
Community with partner 168 (44) 193 (46) 185 (46)
Community with other 54 (14) 55 (13) 54 (13)
Hostel 12 (3) 19 (4) 17 (4)
Nursing home 11 (3) 9 (2) 5 (1)
Cognitive function
SPMSQ score, mean (SD) 8.3 (2.1) 8.3 (2.1) 8.4 (2.0)
Participants with intact cognitive function (SPMSQ score ⱖ8) 316 (75) 280 (73) 310 (77)
Faller in previous 6 mo c 210 (55) 247 (58) 212 (53)
Health-related quality of life score, mean (SD)
EQ-5D VAS 58.5 (12.8) 57.6 (12.9) 57.6 (13.7)
EQ-5D Utility (Dolan method −0.59 to 1.0), mean (SD) 0.46 (0.35) 0.39 (0.36) 0.44 (0.35)
Geriatric depression scale out of a possible 15, mean (SD) 6.9 (2.0) 7.1 (2.0) 6.7 (2.0)
EQ-5D mobility item
No limitations 84 (22) 94 (22) 103 (26)
Some limitations 240 (63) 248 (58) 232 (58)
Severe limitations 51 (13) 80 (19) 63 (16)
EQ-5D personal care item
No limitations 169 (44) 155 (37) 165 (44)
Some limitations 167 (44) 204 (48) 181 (45)
Severe limitations 39 (10) 63 (15) 52 (13)
EQ-5D usual activities item
No limitations 93 (24) 92 (22) 106 (26)
Some limitations 153 (40) 165 (39) 148 (37)
Severe limitations 129 (34) 164 (39) 144 (36)
Days, median (IQR), No.
In study (consent to discharge) 11 (5-31) 14 (6-36) 13 (5-32)
In hospital (both study wards and nonstudy wards) 19 (8-44) 23 (8-51) 20 (7-46)
Between admission to hospital and consent 4 (1-12) 4 (1-14) 4 (1-12)
In study on acute wards (only participants who were on acute study wards) 6 (3-11) 7 (4-13) 6 (3-13)
In study on subacute wards (only participants who were on subacute study wards) 28.5 (14-47) 25 (12-49) 26.5 (15-45)

Abbreviations: EQ-5D (formerly EuroQol)26 quality of life instrument; IQR, interquartile range; SPMSQ, Short Portable Mental Status Questionnaire (SPMSQ)24
(10 is the highest possible SPMSQ score); VAS, visual analog scale.
a Unless otherwise indicated, data are reported as number (percentage) of participants.
b Other geriatric management is its own diagnostic code and is not a summation of all other diagnostic categories.
c Fallers are those who experienced 1 or more falls.

the community or a ward or facility not participating in this S.M.). The randomization envelopes were kept in the locked re-
study. search office at each site, and 1 envelope was opened for each par-
ticipant in order of recruitment on completion of the baseline as-
Randomization and Masking sessment by the trained health professionals providing the
intervention at each site (A.-M.H. and S.M.), who were unaware
A computer-generated, random allocation sequence (without per- of the participant’s result from the baseline assessment. The trained
muted blocks) was developed by the principal investigator (T.P.H.), health professionals then provided the materials-only interven-
and the randomly allocated numbers were placed into opaque, tion or the complete program to participants as soon as practi-
consecutively numbered envelopes by 2 investigators (T.P.H. and cable following this random allocation sequence.

ARCH INTERN MED/ VOL 171 (NO. 6), MAR 28, 2011 WWW.ARCHINTERNMED.COM
518

©2011 American Medical Association. All rights reserved.

Corrected on March 28, 2011
Downloaded From: https://jamanetwork.com/ by Khalid Alomari on 07/08/2023
 5162 Admissions to study wards

3536 Not referred

1626 Referred to study

87 Exclusions
061, Medically unwell until discharge, transfer, or death
026, Emotionally distressed
174 Patients approached but did not provide consent
134, Not interested in study
032 Patients did not think they would benefit
008 Patients felt their hearing was too poor
133 Researchers unable to contact family
0 25 Family members approached but did not provide consent
16 Family members felt patient had insufficient English
1206 Randomized communication skills to benefit
10 Family members approached but were not interested

381, Control group 424, Materials only group 401, Complete program group

Baseline assessment Baseline assessment Baseline assessment

17 Partially incomplete 19 Partially incomplete 9 Partially incomplete

No intervention 409, Materials provided 388, Complete program provided

15, Not provided 2 (2-3), Median (IQR) number of 1-to-1 sessions per participant
(9 sudden discharge, 6 medically unwell) 25 (20-36), Median (IQR) number of minutes in 1-to-1 follow-up
13, Not provided (10 sudden discharge, 3 medically unwell)

1 Withdrawal (data available retained in analysis)

381, Discharge and collation 424, Discharge and collation of falls data 401, Discharge and collation of falls data
of falls data

1206, Analyzed

Figure 1. Study flowchart.

Research assistants who approached participants for con- The rate of falls and rate of injurious falls per 1000 patient-
sent also completed the baseline assessments, weekly falls re- days outcomes were compared across groups using Andersen-
views, and discharge assessments and were blind to group al- Gill Cox recurrent events survival analysis with clustering by
location. Data, with mock codes for group allocation (inserted participant and robust variance estimates.27,28 The proportion
by S.M.), were forwarded to the principal investigator (T.P.H.), of patients who incurred 1 or more falls was compared be-
who undertook the study interim analysis and final data analy- tween groups using logistic regression. For these analyses, an
sis procedures. A blinding survey was also distributed to clini- initial model was constructed that included an interaction term
cal staff members (nursing and allied health) caring for par- between the group variables and the dichotomous variable of
ticipants during the final month of the study recruitment period, whether a patient’s admission SPMSQ score was 7 of 10 or less.
asking the staff members which group they believed their pa- Where significant interaction was identified, simple effects were
tients had been allocated to. investigated for participants with intact cognitive function sepa-
rately from those with impaired cognitive function.
Analysis The Andersen-Gill Cox recurrent events survival analysis
approach models data under the assumption of proportional
Falls outcomes were divided into 3 categories: the rate of falls, the hazards. Nelson-Aalen plots displaying the cumulative hazard
proportion of patients who experienced 1 or more falls (fallers), curves for each group were used to investigate this assump-
andtherateofinjuriousfalls.Therateoffallswasmeasuredinevents tion. Where there was graphical evidence of this assumption
per 1000 patient-days. Injurious falls were defined as falls result- being violated, negative binomial regression was used instead.
ing in bruising, laceration, fracture, loss of consciousness, or pa- Statistical power for this study was calculated using 1000
tient reports of persistent pain. All analyses were conducted with bootstrap simulations of patient-level data previously col-
participants in their assigned groups and were adjusted for whether lected from the Australian hospital setting,2 and the results in-
the patient was treated on a subacute ward during the study (given dicated that our experiment would have 80% power to detect
the imbalance between groups in this factor and the impact this a difference between groups in the rate of falls of 30%. This as-
factor has on length of stay and rate of falls).2,10 sumed a sample size of 390 patients per group, a falls rate of

ARCH INTERN MED/ VOL 171 (NO. 6), MAR 28, 2011 WWW.ARCHINTERNMED.COM
519

©2011 American Medical Association. All rights reserved.

Corrected on March 28, 2011
Downloaded From: https://jamanetwork.com/ by Khalid Alomari on 07/08/2023
 Table 2. Between-Group Comparisons on Fall-Related Outcomes

Outcome Control Materials Only Complete Program

Total Sample
Falls/injurious falls/falls resulting in fracture, No. 81/25/2 96/40/2 70/32/1
Falls per 1000 patient-days 9.27 8.61 7.63
Fallers, No. (%) a 54 (14) 56 (13) 44 (11)
Injurious falls per 1000 patient-days 2.86 3.59 3.49
Cognitively Intact Participants
Falls/injurious falls/falls resulting in fracture, No. 46/15/2 61/25/1 25/10/0
Falls per 1000 patient-days 8.72 8.18 4.01
Fallers, No. (%) a 30 (11) 32 (10) 20 (6)
Injurious falls per 1000 patient-days 2.84 3.34 1.60
Cognitively Impaired Participants
Falls/injurious falls/falls resulting in fracture, No. 35/10/0 35/15/1 45/22/1
Falls per 1000 patient-days 10.10 9.47 15.30
Fallers, No. (%) a 24 (24) 24 (22) 24 (26)
Injurious falls per 1000 patient-days 2.89 4.06 7.49

a Fallers are those who experienced 1 or more falls.

15.7 per 1000 patient-days in the group with the higher falls tal radius and first metacarpal in one case and distal ra-
rate, and a 2-tailed alpha of .05. An additional 12 patients per dius and rib in another; and complete program: orbital
group were recruited (additional 3%) to account for potential fossa and C2 vertebra). Pairwise comparisons did not re-
dropouts, creating a per-group size of n=402 (total, n=1206).
veal significant differences between groups overall
(Table 2, Table 3, and Figure 2). Interaction plots be-
Deviation From Published Protocol
tween group allocation and cognitive impairment for each
The published protocol for this trial did not include detail on use fall outcome (Figure 3) and statistical investigation of
of negative binomial regression where the proportional hazards these revealed significant interaction for each falls out-
assumption did not hold, nor did it include adjustment for whether come (P⬍.05).
the participant was treated on a subacute ward during the study. The rate of falls was significantly lower among par-
These modifications were made in light of the distribution of trial ticipants with intact cognitive function and allocated to
data collected. Examination of the interaction effect between in- the complete program group (4.01 falls per 1000 patient-
tervention group and cognitive impairment was not included in days) compared with the rate among similar partici-
the published protocol despite the previously stated intention of
pants allocated to the control and materials-only groups
the authors to examine this interaction.14
(8.72 falls per 1000 patient-days and 8.18 falls per 1000
patient-days, respectively), and the proportion of these
TRIAL REGISTRATION AND ETHICAL APPROVAL patients who became fallers was lower in the complete
This trial was registered with the Australia New Zealand Clini-
program group than in the control group (6% vs 11%).
cal Trials Registry (ACTRN12608000015347) on January 11, The unadjusted number needed to treat with the com-
2008. Ethical clearance was provided by the medical research plete program to prevent 1 patient becoming a faller rela-
ethics committee of the University of Queensland and the hu- tive to the control group was 32.9; and to prevent 1 fall,
man research ethics committees of the Princess Alexandra Hos- it was 15.4. The proportion of cognitively intact partici-
pital and Swan Districts Hospital. pants in the complete program group who fell was sig-
nificantly lower than that in the control group (6% vs
RESULTS 11%). There was a trend toward a reduction in the rate
of injurious falls among cognitively intact participants
Baseline and demographic characteristics of partici- in the complete program group compared with those in
pants allocated to each group are summarized in Table 1. the control group.
Participants in each group were broadly similar, al- Among participants with impaired cognitive function,
though a noticeable difference was evident for the pro- those allocated to the complete program incurred a sig-
portion of participants allocated to each group who were nificantly higher rate of injurious falls per 1000 patient-
recruited from a subacute ward. There were no control days than participants in the control group (7.49 vs 2.89).
participants provided with either of the intervention con- However, there were no serious injuries (fractures) in-
ditions, but some participants allocated to the interven- curred by any of these patients, and the proportion of par-
tion conditions did not receive their intervention for rea- ticipants with impaired cognitive function who fell was com-
sons presented in Figure 1. parable (complete program, 26%; control, 24%).
There were 247 falls across the study sample and 97 The median (interquartile range) time spent by the
injurious falls (Table 2 and Table 3). Five falls re- trained health professional setting up the multimedia ma-
sulted in fractures (control: pubic rami and sacrum in terials and in face-to-face contact with participants in the
one case and olecranon in another; materials only: dis- complete group was 25 (20-36) minutes, with a maxi-

ARCH INTERN MED/ VOL 171 (NO. 6), MAR 28, 2011 WWW.ARCHINTERNMED.COM
520

©2011 American Medical Association. All rights reserved.

Corrected on March 28, 2011
Downloaded From: https://jamanetwork.com/ by Khalid Alomari on 07/08/2023
 Table 3. Between-Group Comparisons on Fall-Related Outcomes a

Materials Only vs Complete Program vs Complete Program vs

Control as Reference Control as Reference Materials Only as Reference

Outcome Ratio P Value Ratio P Value Ratio P Value

Total Sample
Falls per 1000 patient-days 0.91 (0.61-1.36) .65 0.83 (0.54-1.27) .39 0.91 (0.58-1.42) .63
Fallers b 0.84 (0.55-1.27) .40 0.74 (0.48-1.15) .18 0.89 (0.58-1.38) .62
Injurious falls per 1000 patient-days 1.21 (0.67-2.17) .53 1.22 (0.69-2.20) .49 0.99 (0.56-1.76) .99
Cognitively Intact Participants
Falls per 1000 patient-days 0.83 (0.48-1.44) .51 0.43 (0.24-0.78) .006 0.51 (0.28-0.93) .03
Fallers b 0.80 (0.46-1.38) .41 0.51 (0.28-0.94) .03 0.65 (0.36-1.18) .16
Injurious falls per 1000 patient-days 0.96 (0.44-2.08) .92 c 0.53 (0.23-1.22) .13 c 0.55 (0.23-1.27) .16 c
Cognitively Impaired Participants
Falls per 1000 patient-days 0.99 (0.55-1.78) .97 c 1.48 (0.86-2.53) .15 c 1.45 (0.82-2.59) .21 c
Fallers b 0.92 (0.48-1.78) .82 1.38 (0.70-2.75) .35 1.49 (0.75-2.95) .25
Injurious falls per 1000 patient-days 1.51 (0.64-3.57) .35 2.63 (1.19-5.84) .02 1.98 (0.92-4.25) .08

a Unless otherwise indicated, data are reported as adjusted hazard ratios (robust 95% confidence intervals [CIs]) or adjusted odds ratios (95% CIs). All analyses
were adjusted for whether the patient was treated on a subacute hospital ward during the study.
b Fallers are those who experienced 1 or more falls.
c Negative binomial regression incidence rate ratio (95% CI); P value used if proportional hazards assumption violated.

Control group Materials only Complete program

Overall Intact cognitive function Impaired cognitive function

2.00 1.50 3.00
Cumulative Falls per Patient

1.00 2.00
1.00
0.50 1.00

0.00 0.00 0.00

0 50 100 150 200 250 0 50 100 150 200 0 50 100 150 200 250

0.60 0.40 0.80

Cumulative Injurious Falls

0.40
per Patient

0.20 0.40

0.20

0.00 0.00 0.00

0 50 100 150 200 250 0 50 100 150 200 0 50 100 150 200 250
Length of Time in Study, d Length of Time in Study, d Length of Time in Study, d

No. at Risk
Complete program 401 48 11 2 2 0 310 34 4 0 0 091 14 7 2 2 0
Materials only 424 71 12 5 1 0 316 44 8 3 0 108 27 4 2 1 0
Control group 381 42 14 4 1 0 280 22 7 0 0 101 20 7 4 1 0

Figure 2. Nelson-Aalen cumulative hazard curves for rates of falls and rates of injurious falls outcomes.

mum of 200 minutes for 1 participant. Of the 280 pa- wear or clothing (38 patients), and doing more exercise
tients allocated to the complete program group at the Prin- to get stronger and better balance (34 patients). Of the
cess Alexandra Hospital site, 273 patients recorded a total 299 patients allocated to the materials-only interven-
of 700 goals in relation to behavior modification in their tion at the Princess Alexandra Hospital site, 31 patients
education materials. The most common goal (142 pa- recorded a total of 75 goals. The most common goals re-
tients) related to asking for help, followed by identify- lated to asking for help and waiting for help to arrive once
ing environmental hazards (131 patients), using walk- it had been asked for (14 patients each), followed by iden-
ing aids or other aids (97 patients), waiting for help after tifying environmental hazards (9 patients) and using aids
it has been asked for (71 patients), wearing safe foot- (8 patients).

ARCH INTERN MED/ VOL 171 (NO. 6), MAR 28, 2011 WWW.ARCHINTERNMED.COM
521

©2011 American Medical Association. All rights reserved.

Corrected on March 28, 2011
Downloaded From: https://jamanetwork.com/ by Khalid Alomari on 07/08/2023
 subgroup analysis revealed that the effect of the com-
A Control
plete program (with trained health professional follow-
18 Materials only up) was modified by whether the patient had cognitive
Falls per 1000 Participant Days, No.
Complete program impairment for each of the 3 falls outcomes examined.
15
For cognitively intact patients, the complete program pro-
12
duced a relatively consistent and sizeable reduction
(~50%) across each of the falls outcomes examined, and
9 the difference was significant in 2 of these outcomes. The
complete program also demonstrated a significant re-
6 duction in rate of falls among cognitively intact partici-
3
pants relative to the materials-only group. The magni-
tude of reduction in falls outcomes was comparable with
0 results from previous research examining an earlier ver-
Intact Impaired
Cognitive Function
sion of this intervention in the hospital setting.14 Hence,
there is now growing evidence that the complete pro-
B
gram may be an effective strategy for preventing falls
30
among hospital patients who are cognitively intact.
25 Many of the strategies pursued by patients as a result of
participating in the complete program focused on (1) work-
20 ing more effectively with staff members caring for them;
Fallers, %

(2) identifying environmental hazards; and (3) using ap-

15
propriate aids, equipment, and clothing. These proposed
10 strategies form a plausible mechanism of action for reduc-
ing falls among these patients and highlight the impor-
5 tance of behavioral elements in the causes of falls in this
setting. However, the complete program was not an effec-
0
Intact Impaired tive strategy and may even be harmful for patients with im-
Cognitive Function paired cognitive function: the rate of injurious falls was
C higher in this group. Cognitive impairment can limit the
ability of patients to adhere to the planned safety-
Injurious Falls per 1000 Participant Days, No.

9
promoting behaviors and is a reason why an education pro-
gram might not be beneficial among these patients. How-
6
ever, reasons why it may be harmful are less apparent. It is
possible that the education process made these partici-
pants more willing to report injuries from falls, such as pain,
to the blinded research assistants. In support of this no-
3
tion, we found a discrepancy in the proportion of injuri-
ous falls to total falls reported by patients in the materials-
only (43% of falls were injurious) and complete program
0 (49% injurious) groups compared with those in the con-
Intact Impaired
Cognitive Function
trol group (29% injurious).
Our study was limited by its inability to conceal from
study participants their group allocations, although this
Figure 3. Interaction plots between group allocation and cognitive function
(intact/impaired) on falls outcomes. limitation is common for education-based interven-
tions. It may have influenced results because partici-
pants allocated to the intervention groups may have been
The cross-sectional hospital staff blinding survey re-
particularly motivated to avoid falls by virtue of know-
vealed that of 54 study participants, only 16 had their
ing they had been allocated to the intervention groups.
group allocation correctly identified by their primary care
This enhanced level of motivation might not have been
nurse (29%) (␬ = −0.05), and 17 had their group cor-
present had the complete program been provided out-
rectly identified by their treating physiotherapist (31%)
side the research context.
(␬=−0.06). No adverse events were reported directly from
The simple randomization approach used in the pres-
interaction with the education materials.
ent study generated groups that were not equal in total
size or proportion recruited from subacute hospital wards.
COMMENT The investigators had anticipated that a simple random-
ization approach would be sufficient for generating groups
The 2 models of patient education did not significantly of relatively equal size and comparable baseline demo-
reduce falls outcomes across the entire sample. This study graphics given the number of participants being re-
was one of the few falls prevention randomized trials to cruited. However, the discrepancy in proportion of par-
specifically target cognitively impaired patients for re- ticipants recruited from subacute hospital wards
cruitment, and this decision was made on the basis of en- necessitated adjustment for this in the analyses to ac-
couraging findings from an earlier trial. An exploratory count for the effect of this imbalance.

ARCH INTERN MED/ VOL 171 (NO. 6), MAR 28, 2011 WWW.ARCHINTERNMED.COM
522

©2011 American Medical Association. All rights reserved.

Corrected on March 28, 2011
Downloaded From: https://jamanetwork.com/ by Khalid Alomari on 07/08/2023
 A high proportion of patients admitted to study wards (Dr Hill); Institute of Health and Biomedical Innovation
were not approached for consent to enter this study. Study and School of Public Health, Queensland University of
recruiters were not present 7 days per week at partici- Technology, and Centre for Functioning and Health Re-
pating sites, and periods of leave meant that not all pa- search, Queensland Health, Brisbane (Mr McPhail); School
tients admitted could be approached before they were of Community and Health Sciences, City University, Lon-
within 3 days of anticipated discharge, particularly on don, England (Dr Oliver); and Western Australia Centre
acute wards. for Health and Ageing, Centre for Medical Research, and
The present study has strengths relative to previous work School of Medicine and Pharmacology, University of West-
in this field. Most importantly, this study has evaluated the ern Australia, Perth, Western Australia, Australia (Dr Beer).
efficacy of a single intervention for the prevention of falls Correspondence: Terry P. Haines, PhD, Allied Health
in hospitals. Previous studies that have demonstrated a re- Clinical Research Unit, Physiotherapy Department,
duction in falls in this setting have all used multifactorial Monash University and Southern Health, Kingston Cen-
interventions, and it was very difficult to determine which tre, Kingston Road, Cheltenham, Victoria 3192, Austra-
elements were the most important.2,9,11 lia ([email protected]).
The present study has not only investigated patient Author Contributions: Dr Haines had full access to all of
education in isolation, but it has analyzed 2 models of the data in the study and takes responsibility for the integ-
patient education so that the important elements within rity of the data and the accuracy of the data analysis. Study
this approach can be further identified. As a result, we concept and design: Haines, A.-M. Hill, K. D. Hill, McPhail,
now know that low-cost, materials-only educational ap- Oliver, Brauer, and Hoffmann. Acquisition of data: Haines,
proaches are unlikely to be of benefit. A.-M. Hill, McPhail, and Beer. Analysis and interpretation
This study also used the most rigorous approach to of data: Haines, A.-M. Hill, K. D. Hill, Oliver, Hoffmann,
collection of falls data (primary outcome) in a random- and Beer. Drafting of the manuscript: Haines, McPhail, and
ized trial to date. This is particularly important because Oliver. Critical revision of the manuscript for important in-
falls prevention research in hospitals commonly relies on tellectual content: Haines, A.-M. Hill, K. D. Hill, McPhail,
reports from third parties (hospital clinicians) who are Oliver, Brauer, Hoffmann, and Beer. Statistical analysis:
commonly not blinded to the research hypothesis or par- Haines and A.-M. Hill. Obtained funding: Haines, K. D. Hill,
ticipant group allocation and who have been shown to and Brauer. Administrative, technical, and material support:
be inconsistent in their approach to classifying and re- Haines, A.-M. Hill, K. D. Hill, McPhail, Brauer, Hoff-
porting falls.20,22 In the present study, a research assis- mann, and Beer. Study supervision: Haines, A.-M. Hill,
tant blinded to participant group allocation collated data McPhail, Oliver, Brauer, Hoffmann, and Beer.
not only from medical records and computerized inci- Financial Disclosure: None reported.
dent reports but also from direct, weekly patient inter- Funding/Support: This trial was supported by National
views. It was impossible to blind hospital staff to partici- Health and Medical Research Council (Australia)
pant group allocation in the present study, but the surveys (NHMRC) project grant number 456097 (Drs Haines, Hill,
of hospital staff members revealed that they were largely Brauer, Oliver, and Hoffmann), an NHMRC Career De-
unaware of participant group allocation. velopment Award (Dr Haines), and a Menzies Research
Further research is warranted to examine the effi- Scholarship (Ms Hill).
cacy of the complete program targeted at cognitively in-
tact patients and used within the context of a broader falls-
REFERENCES
prevention program that uses other strategies to reduce
falls among cognitively impaired patients. Such an in-
1. US Department of Veterans Affairs. National Center for Patient Safety 2004 Falls
tervention may need to take the form of a complex in- Toolkit. http://www4.va.gov/ncps/SafetyTopics/fallstoolkit/index.html. Ac-
tervention that can adapt to the specific strengths and cessed April 2009.
limitations of individual wards. Even with the educa- 2. Haines TP, Bennell KL, Osborne RH, Hill KD. Effectiveness of targeted falls pre-
tion intervention investigated in the present trial, the cost- vention programme in subacute hospital setting: randomised controlled trial. BMJ.
2004;328(7441):676-679.
effectiveness of this approach is likely to vary between 3. Aditya BS, Sharma JC, Allen SC, Vassallo M. Predictors of a nursing home place-
acute and subacute wards owing to the higher rate of falls ment from a non-acute geriatric hospital. Clin Rehabil. 2003;17(1):108-113.
and slower throughput on subacute wards; thus, the cost- 4. Oliver D, Killick S, Even T, Willmott M. Do falls and falls-injuries in hospital indicate
effectiveness of this and future interventions should also negligent care—and how big is the risk? a retrospective analysis of the NHS Liti-
be examined. gation Authority Database of clinical negligence claims, resulting from falls in hos-
pitals in England 1995 to 2006. Qual Saf Health Care. 2008;17(6):431-436.
5. Haines TP, Bell RA, Varghese PN. Pragmatic, cluster randomized trial of a policy
Accepted for Publication: July 31, 2010. to introduce low-low beds to hospital wards for the prevention of falls and fall
Published Online: November 22, 2010. doi:10.1001 injuries. J Am Geriatr Soc. 2010;58(3):435-441.
/archinternmed.2010.444 6. Mayo NE, Gloutney L, Levy AR. A randomized trial of identification bracelets to
prevent falls among patients in a rehabilitation hospital. Arch Phys Med Rehabil.
Author Affiliations: Allied Health Clinical Research Unit, 1994;75(12):1302-1308.
Physiotherapy Department, Monash University and South- 7. Tideiksaar R, Feiner CF, Maby J. Falls prevention: the efficacy of a bed alarm sys-
ern Health, Cheltenham, Victoria, Australia (Dr Haines); tem in an acute-care setting. Mt Sinai J Med. 1993;60(6):522-527.
Physiotherapy Department (Ms Hill and Dr Brauer) and 8. Donald IP, Pitt K, Armstrong E, Shuttleworth H. Preventing falls on an elderly
Division of Occupational Therapy (Dr Hoffmann), Uni- care rehabilitation ward. Clin Rehabil. 2000;14(2):178-185.
9. Healey F, Monro A, Cockram A, Adams V, Heseltine D. Using targeted risk factor
versity of Queensland, Brisbane, Queensland, Australia; reduction to prevent falls in older in-patients: a randomised controlled trial. Age
LaTrobe University, Northern Health and the National Age- Ageing. 2004;33(4):390-395.
ing Research Institute, Melbourne, Victoria, Australia 10. Cumming RG, Sherrington C, Lord SR, et al; Prevention of Older People’s Injury

ARCH INTERN MED/ VOL 171 (NO. 6), MAR 28, 2011 WWW.ARCHINTERNMED.COM
523

©2011 American Medical Association. All rights reserved.

Corrected on March 28, 2011
Downloaded From: https://jamanetwork.com/ by Khalid Alomari on 07/08/2023
 Falls Prevention in Hospitals Research Group. Cluster randomised trial of a tar- 19. World Health Organization. Definition of a fall. www.who.int/violence_injury
geted multifactorial intervention to prevent falls among older people in hospital. _prevention/other_injury/falls/en/index.html. Accessed January 15, 2010.
BMJ. 2008;336(7647):758-760. 20. Haines TP, Massey B, Varghese P, Fleming J, Gray L. Inconsistency in classifi-
11. Stenvall M, Olofsson B, Lundström M, et al. A multidisciplinary, multifactorial cation and reporting of in-hospital falls. J Am Geriatr Soc. 2009;57(3):517-
intervention program reduces postoperative falls and injuries after femoral neck 523.
fracture. Osteoporos Int. 2007;18(2):167-175. 21. Sari AB, Sheldon TA, Cracknell A, Turnbull A. Sensitivity of routine system for
12. Cameron ID, Murray GR, Gillespie LD, et al. Interventions for preventing falls in reporting patient safety incidents in an NHS hospital: retrospective patient case
older people in nursing care facilities and hospitals. Cochrane Database Syst Rev. note review. BMJ. 2007;334(7584):79.
2010;(1):CD005465. doi:10.1002/14651858.CD005465.pub2. 22. Haines TP, Cornwell P, Fleming J, Varghese P, Gray L. Documentation of in-
13. Campbell AJ, Robertson MC. Rethinking individual and community fall preven- hospital falls on incident reports: qualitative investigation of an imperfect process.
tion strategies: a meta-regression comparing single and multifactorial interventions. BMC Health Serv Res. 2008;8:254.
Age Ageing. 2007;36(6):656-662. 23. Hill AM, Hoffmann T, Hill K, et al. Measuring falls events in acute hospitals-a com-
14. Haines TP, Hill KD, Bennell KL, Osborne RH. Patient education to prevent falls in parison of three reporting methods to identify missing data in the hospital re-
subacute care. Clin Rehabil. 2006;20(11):970-979. porting system. J Am Geriatr Soc. 2010;58(7):1347-1352.
15. Hill AM, Hill K, Brauer S, et al. Evaluation of the effect of patient education on 24. Pfeiffer E. A short portable mental status questionnaire for the assessment of
rates of falls in older hospital patients: description of a randomised controlled organic brain deficit in elderly patients. J Am Geriatr Soc. 1975;23(10):433-
trial. BMC Geriatr. 2009;9:14. 441.
16. Janz NK, Becker MH. The Health Belief Model: a decade later. Health Educ Q. 25. Sheikh JI, Yesavage JA, Brooks JO III, et al. Proposed factor structure of the
1984;11(1):1-47. Geriatric Depression Scale. Int Psychogeriatr. 1991;3(1):23-28.
17. Hill AM, McPhail S, Hoffmann T, et al. A randomized trial comparing digital video 26. Rabin R, de Charro F. EQ-5D: a measure of health status from the EuroQol Group.
disc with written delivery of falls prevention education for older patients in hospital. Ann Med. 2001;33(5):337-343.
J Am Geriatr Soc. 2009;57(8):1458-1463. 27. Therneau TM, Hamilton SA. rhDNase as an example of recurrent event analysis.
18. Haines T, Kuys SS, Morrison G, Clarke J, Bew P. Cost-effectiveness analysis of Stat Med. 1997;16(18):2029-2047.
screening for risk of in-hospital falls using physiotherapist clinical judgement. 28. Robertson MC, Campbell AJ, Herbison P. Statistical analysis of efficacy in falls
Med Care. 2009;47(4):448-456. prevention trials. J Gerontol A Biol Sci Med Sci. 2005;60(4):530-534.

Call for Photographs

The Archives is seeking photographs to be included as

fillers in our journal. We believe that our readers may
be an excellent source of interesting and thoughtful pho-
tographs. If you would like us to consider your photog-
raphy for publication, we invite you to submit your pho-
tograph to our Web-based submission site under the
category Images From Our Readers at http://manuscripts
.archinternmed.com. Please upload photograph submis-
sions in .jpg or .tif format. Hard copy photographs are
not acceptable. For more information please e-mail
[email protected].

ARCH INTERN MED/ VOL 171 (NO. 6), MAR 28, 2011 WWW.ARCHINTERNMED.COM
524

©2011 American Medical Association. All rights reserved.

Corrected on March 28, 2011
Downloaded From: https://jamanetwork.com/ by Khalid Alomari on 07/08/2023