Sei sulla pagina 1di 57
Home safety in falls prevention: A closer look at the CHEP programme Li Ruijie, MSc
Home safety in falls prevention:
A closer look at the CHEP programme
Li Ruijie, MSc OT
Senior Research Analyst
HSOR
The evidence behind your decisions
Contents  Introduction  Research design  Results  Discussion  Conclusion
Contents
 Introduction
 Research design
 Results
 Discussion
 Conclusion
Introduction
Introduction
Introduction: CHEP  The Community Health Engagement Programme (CHEP) aims to engage the seniors in
Introduction:
CHEP
 The Community Health Engagement Programme
(CHEP) aims to engage the seniors in an active and
fulfilling lifestyle.
 The objectives are:
- Equip residents in the community with skills to keep them
in the community.
- Integrate Tan Tock Seng Hospital (TTSH) with community
services within specific geographic locations to ensure
smooth transition of care.
Introduction: "Stepping Out"  "Stepping Out" is the falls prevention component of CHEP. 
Introduction:
"Stepping Out"
 "Stepping Out" is the falls prevention component of CHEP.
 Components of the "Stepping Out" include:
- Screening for geriatric problems.
- Exercises to improve gait and strength.
- Home visits to
> Assess home safety.
> Monitor medication compliance.
- Health talks on home safety and dietary needs.
- Training of volunteers to give assistance and guidance to elders.
Introduction: "Stepping Out"  Several outcome measures collected within the "Stepping Out" programme
Introduction:
"Stepping Out"
 Several outcome measures collected within the "Stepping
Out" programme
- Falls
- Berg’s Balance Scale (BBS)
- Falls Efficacy Scale (FES)
- Refactored SAFER HOME version 3
 These outcome measures allow us to better understand
the factors associated with falls.
Introduction: Research question What are the factors associated with falls within the "Stepping Out" programme?
Introduction:
Research question
What are the factors associated with falls
within the "Stepping Out" programme?
Research design
Research design
Research design: Enrolment  Potential participants were - Selected from predetermined precincts. - Assessed by
Research design:
Enrolment
 Potential participants were
- Selected from predetermined precincts.
- Assessed by physiotherapists and occupational therapists.
 Participants with moderate to high risk of falls were
enrolled for a year.
 For the purposes of this study, only the first 3 months
data are used.
Research design: Outcome measures  Falls - Number of falls  Falls Efficacy Scale (FES)
Research design:
Outcome measures
 Falls
- Number of falls
 Falls Efficacy Scale (FES)
- A scale to measure the fear of falling.
- (Higher score implies more confidence)
Research design: Outcome measures  Berg’s Balance Scale (BBS) - A scale to measure a
Research design:
Outcome measures
 Berg’s Balance Scale (BBS)
- A scale to measure a person’s ability to balance.
- (Higher score implies better balance)
 SAFER HOME
- An instrument designed to measure home safety.
- (Higher score implies worse home safety)
Outcome measures: SAFER HOME  The SAFER HOME was refactored to provide alternative insights into
Outcome measures:
SAFER HOME
 The SAFER HOME was refactored to provide alternative
insights into the concept of home safety.
 Various sources of inputs and evidence provide validity
to the refactored instrument.
- Theoretical model (Model of Human Occupation).
- Expert opinion from occupational therapists.
- Statistical justifications using confirmatory factor analysis.
Outcome measures: SAFER HOME Original factors New factors 1. Bathroom & Toilet 1. Environment 2.
Outcome measures:
SAFER HOME
Original factors
New factors
1. Bathroom & Toilet
1. Environment
2. Communication & Scheduling
2. Performance capacity
3. Eating
4. Environmental hazards
5. Household
6. Kitchen
7. Leisure
8. Living situation
9. Medication, addiction & abuse
10. Mobility
11. Personal care
12. Wandering
Outcome measures: SAFER HOME  Refactored instrument consist of 2 factors: - Environment > Defined
Outcome measures:
SAFER HOME
 Refactored instrument consist of 2 factors:
- Environment
> Defined as items amendable to environment modification.
- Performance capacity
> Defined as items amendable to remediation of different domains
of the participant (e.g. physical, cognitive etc.)
Research design: Outcome measures Follow-up measures • Falls Month 1 Month 2 Month 3 Baseline
Research design:
Outcome measures
Follow-up measures
• Falls
Month 1
Month 2
Month 3
Baseline measures
Follow-up measures
• Falls
• BBS
• BBS
• FES
• FES
• SAFER HOME
• SAFER HOME
Research design: Outcome measures  Change scores were computed for the following measures: - BBS
Research design:
Outcome measures
 Change scores were computed for the following
measures:
- BBS
- FES
- SAFER HOME
 Change score is computed by: Month 3 – Month 1
- A positive score indicates an increase in the score from
baseline.
Research design: Analysis framework  "Generalised Linear Model" (GLM) was used to model the variables.
Research design:
Analysis framework
 "Generalised Linear Model" (GLM) was used to model
the variables.
 Dependent variable
- The number of falls in the follow-up period.
 Independent variables
- Change scores of the various assessments.
 Covariates
- Demographic variables and baseline characteristics.
Research design: Analysis framework  As the dependent variable is "falls" which is the count
Research design:
Analysis framework
 As the dependent variable is "falls" which is the count of falls,
Poisson regression is typically used.
 2 problems often arise in the analysis of count data:
- Over dispersion
- Excessive zeroes
 May need to consider the use of
- Zero-inflated models
- Negative binomial regression
- Both
Research design: Analysis framework  Given the large number of zeroes in "falls", a zero-
Research design:
Analysis framework
 Given the large number of
zeroes in "falls", a zero-
inflated model would be
adopted.
 A zero-inflated negative
binomial model was
eventually adopted as the
model of choice for
analysing the data.
Results
Results
Results: Descriptive statistics Sample size 371 Age, Mean (SD) 73.086 ( 7.846) Gender, Count (%)
Results:
Descriptive statistics
Sample size
371
Age, Mean (SD)
73.086 ( 7.846)
Gender, Count (%)
Female
274
(73.854)
Ethnicity , Count (%)
Chinese
309
(83.288)
Malay
44
(11.860)
Indian
13
( 3.504)
Others
5 ( 1.348)
Results: Modelling diagnostics  The negative binomial distribution was adopted to account for possible over
Results:
Modelling diagnostics
 The negative binomial distribution was adopted to
account for possible over dispersion after accounting for
the excess zeroes.
 The Vuong test was used to determine which is the
superior model (zero-inflated vs. non zero-inflated)
- Test-Statistic = 4.200, p < .05
- Test results suggest that the zero-inflated model is a more
suitable model.
Results: Modelling results (Demographics) Estimate Std. Error z value Pr(>|z|) Age -0.110 0.030 -3.716 0.000
Results:
Modelling results (Demographics)
Estimate
Std. Error
z value
Pr(>|z|)
Age
-0.110
0.030
-3.716
0.000
*
Gender (Female)
-1.736
0.517
-3.358
0.001
*
Ethnicity (Malay)
0.054
0.502
0.107
0.915
Ethnicity (Indian)
2.674
1.068
2.505
0.012
*
Ethnicity (Others)
0.573
1.133
0.506
0.613
Results: Modelling results (Baseline) Estimate Std. Error z value Pr(>|z|) Falls 1.237 0.296 4.174 0.000
Results:
Modelling results (Baseline)
Estimate
Std. Error
z value
Pr(>|z|)
Falls
1.237
0.296
4.174
0.000
*
BBS
-0.135
0.042
-3.203
0.001
*
FES
0.466
0.228
2.046
0.041
*
Environment
0.118
0.135
0.876
0.381
Performance capacity
0.091
0.110
0.820
0.412
Results: Modelling results (Change scores) Estimate Std. Error z value Pr(>|z|) BBS -0.276 0.070 -3.972
Results:
Modelling results (Change scores)
Estimate
Std. Error
z value
Pr(>|z|)
BBS
-0.276
0.070
-3.972
0.000
*
FES
-0.005
0.176
-0.027
0.979
Environment
0.330
0.165
-1.993
0.046
*
Performance capacity
-0.128
0.162
-0.787
0.431
Results: Modelling results (Significant variables) IRR Demographic Age 0.896 Gender (Female) 0.176 Ethnicity
Results:
Modelling results (Significant variables)
IRR
Demographic
Age
0.896
Gender (Female)
0.176
Ethnicity (Indian)
14.502
Baseline
Falls
3.445
BBS
0.873
FES
1.593
Change scores
BBS
0.759
Environment
1.391
Results: Modelling results (Zero model)  None of the variables were significant in the zero
Results:
Modelling results (Zero model)
 None of the variables were significant in the zero model.
- Zeroes were not well modelled.
- However the Vuong test indicates that the zero-inflated
model is superior to the non zero-inflated model.
Discussion
Discussion
Discussion (Demographics)  Older people have lower number of falls.  Females have lower number
Discussion (Demographics)
 Older people have lower number of falls.
 Females have lower number of falls than males.
 Chinese have a lower number of falls than Indians
Discussion (Baseline)  People with lower number of falls at baseline have a lower number
Discussion (Baseline)
 People with lower number of falls at baseline have a
lower number of falls at follow-up.
 People with better balance have lower number of falls.
 People with more fear of falling have a lower number of
falls.
Discussion (Change scores)  People with an improvement in balance have a lower number of
Discussion (Change scores)
 People with an improvement in balance have a lower
number of falls.
 People with better home safety in "Environment" have a
lower number of falls.
Discussion  2 findings run contrary to established literature: - Older age associated with lower
Discussion
 2 findings run contrary to established literature:
- Older age associated with lower number of falls.
- More fear of falling was associated with lower number of falls.
 Possibly related to different stages of improvement in function.
- With decreased participation as an intermediary.
 Requires further work to understand the trajectory of
improvement in relation to falls.
Discussion  Performance capacity was not significant may be due to a lack of sustainability
Discussion
 Performance capacity was not significant may be due to a
lack of sustainability in change.
 A change in items susceptible to environmental
modification suggests that environmental modification
may be an effective way to reduce falls.
Conclusion  Several factors associated with less falls were uncovered.  2 of the factors
Conclusion
 Several factors associated with less falls were uncovered.
 2 of the factors (age and FES) have results that run
contrary to established literature and a possible
explanation is proposed.
 The refactored SAFER HOME suggests that
environmental modifications has a significant impact on
falls.
Acknowledgements  I would like to thank my collaborators for giving me the opportunity to
Acknowledgements
 I would like to thank my collaborators for giving me the
opportunity to work on this problem and providing
insightful feedback on the interpretation of analysis.
- Ms Ng Lip Chin (TTSH OT Department)
- Dr Noor Hafizah (TTSH Geriatrics Department)
 I would also like to thank Professor John Fox (McMaster
University) for generously providing advice on resolving
some of the issues I had with analysis.
Refactoring of the SAFER HOME Li Ruijie, MSc OT Senior Research Analyst HSOR The evidence
Refactoring of the SAFER HOME
Li Ruijie, MSc OT
Senior Research Analyst
HSOR
The evidence behind your decisions
Introduction  Original SAFER HOME conceived using Person- Environment-Occupation (PEO) model. Occupation 
Introduction
 Original SAFER HOME
conceived using Person-
Environment-Occupation
(PEO) model.
Occupation
 Influence of PEO can be
seen in the subscales of the
instrument (e.g. Kitchen,
Personal care etc.)
Person
Environment
Introduction  Other models exist that can possibly conceptualise the different concepts of home safety.
Introduction
 Other models exist that can possibly conceptualise the
different concepts of home safety.
 Using a sound theoretical model and tapping on the
experience of therapists working on the ground, it is
possible to develop an alternate grouping of the
subscales of the SAFER HOME.
 Adopting an alternate model can offer a new perspective
on the constructs revolving falls and home safety.
Introduction  The model of interest is the “Model of Human Occupation” (MOHO).  The
Introduction
 The model of interest is the “Model of Human
Occupation” (MOHO).
 The primary building blocks of the model are:
- Volition
- Habituation
- Performance capacity
Introduction VOLITION Interests Environment Values Personal causation Participation HABITUATION Occupational
Introduction
VOLITION
Interests
Environment
Values
Personal causation
Participation
HABITUATION
Occupational
Identity
Habits
Occupational
Performance
Adaptation
Roles
Occupational
Competence
Skill
PERFORMANCE
CAPACITY
Introduction  For this study, we would be focusing on - Performance capacity > Defined
Introduction
 For this study, we would be focusing on
- Performance capacity
> Defined as the capacity in the various functional domains
(physical, cognitive etc.) to perform various tasks.
- Environment
> Defined as the environment within which tasks are performed.
Introduction  The aim of this study is therefore to - Use MOHO and expert
Introduction
 The aim of this study is therefore to
- Use MOHO and expert knowledge to regroup the items
within the SAFER HOME
- Use the regrouping to gain additional insights into
relationship of falls with home safety.
Research design
Research design
Research design: Refactoring  The items within SAFER HOME would refactored using the basic elements
Research design:
Refactoring
 The items within SAFER HOME would refactored using the basic
elements of MOHO:
- Environment
- Performance capacity
 The items would be reframed by considering if it was modifiable
by
- environment modification or
- performance capacity changes.
 This is a highly contextual way of regrouping the items in a bid to
find a “local theory”.
Research design: Refactoring  Volition & habituation were not considered as the premise was to
Research design:
Refactoring
 Volition & habituation were not considered as the
premise was to select components for refactoring that
were relevant to the context of “Stepping Out”.
 Volition & habituation deemed to be not easily
observable or measured, much less influenced within this
context.
Research design: Refactoring  Surveyed 5 occupational therapists (average 5.18 years of experience, > 2
Research design:
Refactoring
 Surveyed 5 occupational therapists (average 5.18 years of
experience, > 2 years in “Stepping Out”)
 The OTs were asked to classify each item to determine if it is
more amendable to:
- Environmental modification
- Performance capacity modification
- Both
 An additional option was available if the OTs deem the item
unnecessary and could be deleted.
Research design: Refactoring  The reclassification was tallied and the highest votes wins.  A
Research design:
Refactoring
 The reclassification was tallied and the highest votes
wins.
 A series of rules were used to break a tie. In general:
- Items would be retained where possible.
- Items split between categories would go into both.
 Using the rules, there was no tie unbroken.
Research design: Refactoring analysis  After reclassifying the items, it is necessary to examine if
Research design:
Refactoring analysis
 After reclassifying the items, it is necessary to examine if
the data fits the new structure.
- Confirmatory factor analysis (CFA) used.
- CFA would determine if the refactored structure is
comparable to the old factor structure in terms of data fit.
Research design: Refactoring analysis  If refactored structure is found comparable to the original factor
Research design:
Refactoring analysis
 If refactored structure is found comparable to the original
factor structure, regression modelling would proceed similar to
the previous study.
 Refactored scores would be calculated and used in the
modelling.
 Similar to the previous study:
- Dependent variable: Falls
- Independent variable: Change scores
- Covariates: Demographics & baseline measures
Research design: Enrolment  A total of 928 complete administration of the SAFER HOME was
Research design:
Enrolment
 A total of 928 complete administration of the SAFER
HOME was used to conduct the CFA.
 These were conducted during the course of the “Stepping
out” programme.
Results
Results
Results: Refactoring Original categories No. of items Refactored categories No. of items Bathroom & Toilet
Results:
Refactoring
Original categories
No. of items
Refactored categories
No. of items
Bathroom & Toilet
11
Environment
22
Communication & Scheduling
3
Performance capacity
32
Eating
2
Both
16
Environmental hazards
13
Delete
4
Household
9
Kitchen
8
Leisure
1
Living Situation
3
Medication addiction & Abuse
3
Mobility
10
Personal Care
8
Wandering
1
Results: Confirmatory factor analysis (Path diagram) Error I01 I01 Living situation Error I02 I02 .
Results:
Confirmatory factor analysis (Path diagram)
Error I01
I01
Living situation
Error I02
I02
.
.
.
.
.
.
.
.
Error I72
I72
Wandering
Results: Confirmatory factor analysis (Path diagram) Error I01 I01 Environment Error I02 I02 . .
Results:
Confirmatory factor analysis (Path diagram)
Error I01
I01
Environment
Error I02
I02
.
.
.
.
Performance
Error I72
I72
capacity
For illustration purposes only, the items do not correspond to assigned categories.
Results: Confirmatory factor analysis (Fit indices)  The fit indices suggest that the data fit
Results:
Confirmatory factor analysis (Fit indices)
 The fit indices suggest that the data fit the models well.
RMSEA
SRMR
Recommended
(Hu & Bentler, 1999)
<.060
<.080
Original structure
.049
.056
Refactored structure
.050
.057
Discussion
Discussion
Discussion  The results suggest that the model fit of the 2 models are comparable.
Discussion
 The results suggest that the model fit of the 2 models are
comparable.
 While the refactored model may not have a better fit than
the original model, the primary aim of the refactoring
was to offer different insights into the home safety
phenomenon.