SEM and
rehabilitation
Peek,
M. Kristen PhD. Structural Equation Modeling and Rehabilitation Research.
American
Journal of Physical Medicine & Rehabilitation. 79(3):301-309, May/June
2000.
Structural
equation modeling (SEM) has been used extensively in the social sciences in the
recent past to model human behavior. SEM has been less used by other
disciplines, including rehabilitation medicine. This article begins by
providing an introduction to structural equation modeling through the
discussion of the definition and basic concepts behind SEM using a conceptual
model relevant to rehabilitation medicine that describes the pathway from
health to disability (the Enabling-Disabling model). The next focus is on
several potential pitfalls of which the researcher needs to be aware when using
SEM. After this discussion, a hypothetical example is presented using
structural equation modeling to evaluate the Enabling-Disabling model. The
article concludes with a review of the advantages and disadvantages of using SEM
for clinical and social science research. Throughout the article, references
are provided for a more detailed examination of SEM. (32 ref)
Comment:
Peek advocates using the SEM with the disablement model and suggests that it
"has successfully been used" (p. 302). Quotes Lawrence and Jette (1996, disentangling) on the need for
"clinical trials" that illuminate causal relationships among
impairment, disability, and the entire disablement process, including
"risk factors" that may change over time.
The article includes Figure 2 (p.304),
which illustrates a structural model in which a quality of life measure rests
at the end of a series of exogenous and endogenous variables.
???Lawrence and Jette (1996) cast doubt on
the existence of a direct relationship between pathology and disability, but I
believe that they did not use SEM. They
indicate that functional limitations or abilities are most directly related to
disability.
Peek has published
examples of use of SEM models (search for his name in this page)
______________________________________________________________________
Mulligan,
S. (1998). Application of structural equation modeling in occupational therapy
research. American Journal of Occupational Therapy, 52, 829-34.
Occupational
Therapy Department, School of Health and Human Sciences, University of New
Hampshire, Hewitt Hall, 4 Library Way, Durham, New Hampshire 03824-3563.
Structural
equation modeling (SEM) is a relatively new research method for analyzing
multivariate data. This article provides an introduction and overview of what
SEM is and identifies some potential uses of this technique for occupational
therapy researchers. The application of SEM in the areas of test construction
and treatment effectiveness is highlighted, some cautionary measures related to
this technique are presented, and resources that provide more detailed and
advanced reading on the subject are included.
Comment: Mulligan has published a study in which she
used CFA for a dyspraxia instrument (see later in this page)
Hepburn, D. A., Deary, I. J., MacLeod, K.
M., & Frier, B. M. (1994). Structural equation modelling of symptoms,
awareness, and fear of hypoglycaemia and personality in patients with
insulin-treated diabetes. Diabetes Care, 17, 1273-1280.
Gold, A. E, Frier, B. M., MacLeod, K. M.,
& Deary, I. J. (1997). A structural equation model for predictors of severe
hypoglycaemia in patients with insulin-dependent diabetes mellitus. Diabetic
Medicine, 14, 309-315.
Hawkins WE, Duncan T. Structural equation
analysis of an exercise/sleep health practices model on quality of life of
elderly persons. Percept Mot Skills 1991;72(3 Pt 1):831-6.
_________________________________________________________
Bartlett,
D.J., & Palisano, R.J. (2000). A multivariate model for determinants of
motor change for children with cerebral palsy. Physical Therapy,
80,598-614.
e-mail:
d.bartlett@uwo.ca. Assistant Professor, School of Physical Therapy, Faculty of
Health Sciences, University of Western Ontario, 1588 Elborn College, London,
Ontario, Canada N6G 1H1
Abstract
The purpose of this article is to describe
the development of a theory- and data-based model of determinants of motor
change for children with cerebral palsy. The dimensions of human functioning
proposed by the World Health Organization, general systems theory, theories of
human ecology, and a philosophical approach incorporating family-centered care
provide the conceptual framework for the model. The model focuses on relationships
among child characteristics (eg, primary and secondary impairments,
personality), family ecology (eg, dynamics of family function), and health care
services (eg, availability, access, intervention options). Clarification of the
complex multivariate and interactive relationships among the multiple child and
family determinants, using statistical methods such as structural equation
modeling, is necessary before determining how physical therapy intervention can
optimize motor outcomes of children with cerebral palsy. We propose that the
development and testing of multivariate models is also useful in physical
therapy research and in the management of complex chronic conditions other than
cerebral palsy. Testing of similar models could provide physical therapists
with support for: (1) prognostic discussions with clients and their families,
(2) establishment of realistic and attainable goals, and (3) interventions to
enhance outcomes for individual clients with a variety of prognostic
attributes. (109 ref)
DMT's
Summary: The article's first part
summarizes SEM, and is less useful in doing so than similar treatments in the
rehabilitation literature (Peek, ; Mulligan,
). The article's second part
describes in detail how the authors constructed theory-driven structural and
measurement models. Their five-factor
structural model is an example of a "sequential longitudinal" model
(MacCallum & Austin, 2000, p. 205).
They also discuss their plans for building a measurement model.
The
authors discuss the advantages of SEM approaches to understanding disablement,
particularly in children with CP. They advocate SEM as a way of deciding
whether impairments like co-contraction are best understood as primary problems
or secondary adaptations. Primary and secondary
impairments can be distinguished in a sequential model by the timing of their
appearance as factors; primary factors appear around the time of diagnosis,
while secondary impairments appear later.
Building and comparing of competing structural models could shed light
on such the identity of particular impairments.
Available
online: http://www.ptjournal.org/PTJournal/June2000/v80n6p598.cfm
__________________________________________________________
Sousa
KH. Chen F.
Assistant
Professor, College of Nursing, Arizona State University, Tempe, AZ.
Conference
Abstract
Health-Related
quality of life theory and structural equation modeling... 34th Annual
Communicating Nursing Research Conference/15th Annual WIN Assembly,
"Health Care Challenges Beyond 2001: Mapping the Journey for Research and
Practice," held April 19-21, 2001 in Seattle, Washington.
Source
Communicating Nursing Research, 34(9):290, 2001 Spring.
Karen
Sousa
http://nursing.asu.edu/facultystaff/sousak.htm
__________________________________________________________
Comment: Studies of cardiac care may be especially
interesting in terms of the types of data challenges they present. Examples:
1.
Riegel
BJ. Dracup KA. Glaser D.
Institution
School of Nursing, San Diego State University,
Sharp HealthCare, CA, USA.
Title
A longitudinal causal model of cardiac
invalidism following myocardial infarction.
Source
Nursing Research. 47(5):285-92, 1998 Sep-Oct.
Abstract
BACKGROUND: Invalidism has been discussed in
the cardiovascular literature for decades. Researchers have studied health
perceptions, emotional distress, and dependency in patients after acute
myocardial infarction in an attempt to understand the phenomenon. However, no
theory of the manner in which these variables interact has been proposed.
OBJECTIVES: Using previous research, a model of invalidism was specified in
which individuals' perceptions that their health is poor lead to emotional
distress and increased dependency. As health perceptions improve over time, emotional
distress and dependency decrease. METHOD: Survey data were collected from 111
men and women 1 and 4 months after a first myocardial infarction and were
tested using structural equation modeling. RESULTS: The model was rejected
using a confirmatory approach (chi2(89) = 141.40; p= .00034). The fit indices,
however, suggested an adequate fit of the model to the data (CFI = .96; NNFI =
.94). CONCLUSION: The conclusion is that the model is reasonable and serves as
a starting point for a theory-based empirical exploration of the invalidism
process.
_________________________________________________________
2.
Riegel
B. Dracup K.
Institution
Sharp Memorial Hosp, San Diego, CA.
Title
Social support and cardiac invalidism
following acute myocardial infarction.
Source
Heart & Lung: Journal of Critical
Care, 21(3):298, 1992 May.
Abstract
At least 25% of acute myocardial infarction
(AMI) patients experience long-term adjustment problems such as cardiac
invalidism defined as a complex of low self-esteem (SE), emotional distress
(ED), poor health perceptions (HP), and interpersonal dependency (ID). A
theoretical model was tested in which social support was hypothesized to
influence cardiac invalidism following AMI. A survey design was used to collect
data from 111 first AMI patients 1 and 4 months following hospital discharge.
The data fit the model well when tested with structural equation modeling with
data obtained at 1 month (comparative fit index (CFI)=.997, p=.42) and with
change scores (CFI=1.0, p=.92). At 1 month SS predicted ID (-.87) and health
outlook (.26). HP predicted ED (-.26) and SE (.33). ED predicted SS desired
(.39). SE predicted ED (-.57). Using change scores, SS predicted ED (.53) and
HP (-.41). SE predicted ED (-.25). HP predicted anger (.20). ID predicted SS
received (-.39). Thus, SS was !
a
significant predictor of cardiac invalidism following AMI. The recovery process
was dynamic at 1 month but less so at 4 months. Interventions designed to
prevent cardiac invalidism should begin during or immediately after hospital
discharge and should focus on augmenting and refining social support.
3.
Fontana
AF. Kerns RD. Rosenberg RL. Colonese
KL.
Support,
stress, and recovery from coronary heart disease: a longitudinal causal model.
Health
Psychology. 8(2):175-93, 1989.
Measures
of support, stress, distress, and cardiac symptoms were obtained from a cohort
of 73 male cardiac patients at hospitalization and at 3, 6, and 12 months
thereafter. Sets of general and alternative hypotheses regarding the direction
of causality among these variables were drawn from the literature on cardiac
rehabilitation, stress, and support. Structural equation modeling was used to
evaluate the stability and duration of these hypotheses over three time-lags. The
results showed strong support for the general hypotheses and minimal support
for the alternative hypotheses. Support ameliorated the subsequent experience
of stress and distress and had opposing effects to these variables on cardiac
symptoms. Support was more influential in the first half of the year than it
was in the second half, however, whereas stress was predominant causally in the
second half. Implications of this pattern for clinical intervention are drawn
and directions for further research are proposed.
____________________________________________________________
Resnick
B. Daly MP. (1998). Predictors of functional ability in
geriatric rehabilitation patients.
Rehabilitation Nursing.
23(1):21-9, 1998 Jan-Feb.
University
of Maryland, Baltimore, USA.
This
study explored the impact that motivation, cognitive status, depression, age,
and physical status have on the functional ability of older adults in a
rehabilitation program. The study was based on a hypothesized model and was
conducted with 200 patients on a geriatric rehabilitation unit. Structural
equation modeling was done to test the hypothesized model. The average age of
participants was 78 years, and the majority were female, Caucasian, unmarried,
and had been admitted for rehabilitation after an orthopedic event. The data
fit the hypothesized model; however, only five paths were significant. Mental
status was a significant predictor of function on admission and a direct and
indirect predictor of function at discharge, and diagnosis and age directly
predicted function at discharge.
_________________________________________________________
Duncan
TE. Stoolmiller M. (1993). Modeling social and psychological
determinants of exercise behaviors via structural equation systems. Research
Quarterly for Exercise & Sport.
64(1):1-16, 1993 Mar.
Oregon
Social Learning Center, Eugene.
Recent
advances in structural modeling techniques allow for the testing of complex
models representing social and behavioral processes. However, most reported
applications in sport and physical activity have been limited to simple models
involving variables measured at a single point in time. Therefore, the purpose
of this article is to demonstrate the use of both cross-sectional and
longitudinal latent variable modeling techniques by examining the relationships
among efficacy cognitions, social support, and the exercise behaviors of
sedentary adults. Results revealed a good fit for the re-specified model,
suggesting the existence of a lagged feedback mechanism in which exercise
behaviors influenced residual change in social support. In turn, efficacy
cognitions appeared to serve a mediational function in the synchronous
relationship between social support and exercise behavior. Findings are
discussed in terms of the utility of structural equation modeling techniques
for und!
erstanding
the complex social and cognitive processes involved in exercise behavior.
_________________________________________________________
SEM and
falling
Preliminary
searches of the literature in Medline do not reveal sources. The area ought to be ripe for investigation.
Lawrence
RH. Tennstedt SL. Kasten LE.
Shih J. Howland J. Jette AM. (1998). Intensity and correlates
of fear of falling and hurting oneself in the next year: baseline findings from
a Roybal Center fear of falling intervention.
Journal of Aging & Health.
10(3):267-86, 1998 Aug.
New
England Research Institutes, USA.
__________________________________________________________
Lawrence
RH. Jette AM. (1996). Disentangling the disablement process. Journals of Gerontology Series
B-Psychological Sciences & Social Sciences, 51B(4):S173-82, 1996 Jul.
(43 ref)
ReneeL%NERI@MCImail.com. New England Research Institutes, 9 Galen
Street, Watertown, MA 02172. Internet:
_________________________________________________________________
Lichtenstein
MJ. Dhanda R. Cornell JE. Escalante
A. Hazuda HP. (2000). Modeling impairment: using the disablement process as a
framework to evaluate determinants of hip and knee flexion. Aging (Milano). 12(3):208-20, 2000 Jun.
Division of Geriatrics and Gerontology,
University of Texas Health Science Center at San Antonio 78284, USA.
lichtenstei@uthscsa.edu
Elders
often present to health care providers with multiple inter-related conditions
that determine an individual's ability to function. The disablement process
provides a generalized sociomedical framework for investigating the complex
pathways from chronic disease to disability. At each stage of the main pathway,
associations may exist among primary physical factors and modifying variables
that ultimately have downstream effects on the progression toward disability.
The purpose of the present analysis is to examine the inter-relationships
between a cohesive set of variables primarily at the level of impairment that
may affect hip and knee flexion range of motion (ROM). The San Antonio
Longitudinal Study of Aging enrolled 833 community dwelling Mexican (MA) and
European American (EA) elders aged 64-78 years between 1992 and 1996. Of these,
647 had complete data from both a home-based and performance-based battery of
assessments for these analyses. Concerning impairments, hip ROM was measured
using an inclinometer, and knee ROM using a goniometer. Pain location and
intensity were assessed using the McGill Pain Questionnaire. Peripheral
vascular disease was assessed using doppler brachial and ankle systolic blood
pressures. Ankle and knee reflexes, and vibratory sensation were assessed by a
standardized neurological examination. As to diseases, diabetes was assessed
using a combination of blood glucose levels and self-report, and arthritis by
self-report. Concerning modifying variables, height and weight were directly
measured and used to calculate BMI. Activity level was assessed with the
Minnesota Leisure Time Questionnaire. Analgesic use was assessed by direct
observation of medications taken within the past two weeks. We used structural
equation modeling to test associations between the variables that were
specified a priori. These analyses demonstrate the central role of BMI as a
determinant of hip and knee flexion ROM. For an increase in level of BMI, the
coefficients [SEM] for changes in levels of hip and knee ROM were -0.38 [0.05]
and -0.26 [0.05], respectively. A higher BMI resulted in lower hip and knee
ROM. BMI was also directly associated with prevalent diabetes (0.10 [0.05]) and
arthritis (0.17 [0.05]). However, after adjustment for BMI, diabetes and
arthritis did not have direct independent associations with either hip or knee
ROM. BMI was also indirectly associated with knee, but not hip, ROM through
paths including lower-leg pain, pain intensity, and neurosensory impairments.
Diabetes had an indirect association with hip, but not knee ROM, through a path
including peripheral vascular disease. In conclusion, BMI is a primary direct
determinant of hip and knee ROM. The paths by which diabetes and arthritis lead
to physical disability may be mediated, in part, at the level of impairment by
BMI's association with joint range of motion. Interventions designed to
decrease the impact of diabetes and arthritis on disability should track
changes in BMI and joint ROM to measure the paths that account for the
intervention's success. The observed associations suggest that interventions
targeted to decrease BMI itself may lead to improved function in part through
improved joint ROM.