Transcript No Slide Title

Invariance of symptoms of mood disorder
in people with temporal lobe epilepsy (TLE)
Stephen Bowden, Rachel Reilly,
Fiona Bardenhagen and Mark Cook.
University of Melbourne & St. Vincent’s Hospital, Melbourne.
INTRODUCTION
To fully represent a theoretical or latent-trait structure of behaviour, it is necessary to model
observed variable intercepts and factor means in addition to the familiar factor loadings,
factor variances and covariances, and residual variances (Meredith, 1993; Widaman &
Reise, 1997). Represented in this way, it is possible to fully examine measurement model
invariance across groups. The centrality of assumptions of invariance for fairness in
testing, and theoretical equivalence of validity patterns are widely acknowledged (Byrne et
al., 1988; Meredith, 1993). Since identification of clinical disorders usually involves
interpretations of patterns of means or patterns of validity correlations, it is important to
examine the assumptions of invariance.
Widaman and Reise (1997) have defined several levels of invariance. Configural
invariance involves the finding of a similar pattern of factors and indicator-variable
assignments. Configural invariance is seen in the type of factor-analysis replication
commonly reported in the clinical literature and, when reported from different samples,
indicates a similar, but not necessarily identical pattern of latent variables structure. In
contrast metric invariance involves successively more restrictive assumptions about the
measurement model, (i) weak metric invariance assumes identical factor loadings, (ii)
strong metric invariance assumes identical observed variable intercepts, and (iii) strict
metric invariance assumes identical reliabilities, across groups. “If strong [metric]
invariance holds, group differences in both means and variances on the latent variables,
which represent the constructs in psychological theories, are reflected in group differences
in means and variances on the measured variables (Widaman & Reise, 1997, p. 295).
In clinical evaluation of psychological adjustment amongst people diagnosed with TLE
there is much speculation about the meaning of psychological symptoms. Some have
suggested that symptoms of distress and poor adjustment should be interpreted as
features of personality and psychopathology specific to people with TLE (Blumer, 1999). In
contrast, others have suggested that exotic formulations of psychopathology distract from
recognition and treatment of familiar psychological disorders, commonly seen in many
people with significant disability (Devinsky, & Souhel, 1999). Central to the debate are
symptoms of depression.
If symptoms of mood disorder in people with TLE reflect behaviours specific to this clinical
population, then it is reasonable to expect that elements of a measurement model of
depression should differ from the model of depressive symptoms seen in other samples. In
contrast, if we were to observe invariance of depressive symptoms across samples, then
this observation would provide compelling evidence for the assumption that symptoms of
mood disorder in people with TLE should be interpreted and treated as depression.
METHOD
Samples: Two samples were employed for this study, the first comprising 187 consecutive
patients undergoing video-telemetry at St. Vincent’s Hospital Melbourne, for the investigation
of TLE. The second sample comprised 150 consecutive patients admitted to the Clinical
Neurosciences Department for investigation and treatment of heterogeneous neurological
disorders, but not seizure disorders. The sample of patients with TLE comprised 94 females
and 93 males, with a mean age of 35.6. The neurological sample comprised 59 females and
91 males, with a mean age of 46.0.
Measure: Depressive symptoms were measured with the Beck Depression Inventory (BDI),
an instrument which is based on a well-validated clinical model of depression (Byrne, 1998;
Tanaka & Huba, 1984). Mean BDI score for the TLE sample was 10.6, and for the neurological
sample 12.8.
RESULTS
Data analysis: Scores on the BDI in both samples were subjected to maximum-likelihood
confirmatory analysis using Lisrel 8.12 replicating, in both sample, the oblique three-factor
solution reported by Byrne (1998), which measure the latent variables of (i) Negative Affect, (ii)
Work/Performance Inhibition, and (iii) Somatic Elements. When this three-factor baseline model
was analysed simultaneously in both groups, using the same method of model identification to
that reported by Widaman & Reise (1997, p. 305) configural invariance was observed and the
resulting fit statistics are reported in Table 1 (Model 1). Next, invariance constraints were placed
on the factor-loading (Λ) matrix, the variable intercept (τ) vector, and the error variance matrix
(Θ), to test the weak, strong, and strict invariance models (Models 2-4), respectively. The
increments in χ2 and values of other goodness of fit statistics reported in Table 1, suggest that
strong invariance obtains, and strict invariance is a plausible assumption.
_______________________________________________________________________________________________
Table 1
Invariance Model
χ² df
Δχ² ΔdF RMSEA
SRMR
CAIC
TLI CFI
_______________________________________________________________________________________________
1 Baseline
619.41
372
.045
.066
1519
.86
642.43
390
23.02 18
.044
.075
1419
.86
.87
669.62
408
27 19 18
.044
.075
1324
.86
.87
708.78
429
39.16 21
.044
.079
1220
.86
.86
.87
(configural invariance)
2 Weak
(Model 1 and Λ invariant)
3 Strong
(Model 2 and τ invariant)
4 Strict
(Model 3 and Θ invariant)
________________________________________________________________________________________________
DISCUSSION
The results of the present study show that a well validated model of depressive symptoms
is invariant across samples of patients with TLE and heterogeneous neurological
disorders. Initial modeling (Model 1), showed configural invariance across samples, which
implies that the same latent variable structure is measured by an identical pattern of items
from the BDI. In other words, it can be assumed that the BDI measures a set of
depressive symptoms in people with TLE, which is the same as that observed in our
neurological reference group, and has been observed in numerous other samples (Byrne,
1998).
In addition, the measurement model displayed precise evidence of strong metric
invariance (Model 3), which implies that the same set of latent variables is measured
across groups, with the same metric. The assumption of strong metric invariance is
necessary for the uncomplicated (scale equivalent) interpretation of patterns of elevated
scores, and external validity patterns (Widaman & Reise, 1997). Overall, this pattern of
strong factorial invariance provides a precise test of the assumption that the behaviours
underlying scores on the BDI are the same in people with TLE, as they are in people with
other neurological conditions, and numerous other community samples (Byrne, 1998). The
results suggest that exotic formulations of personality structure and depressive symptoms
in people with TLE should be avoided, and the symptoms of mood disturbance should be
treated in the same way as in other clinical populations, namely, as potential markers of
depression. Strong validity data, of the kind reported in this study, should sensitise
clinicians to the presence of mood disturbance in patients with organic conditions.
REFERENCES
Blumer, D. (1999). Devinsky, O., Souhel, N. (1999). Evidence supporting the temporal lobe epilepsy personality syndrome. Neurology, 53, S9-12.
Byrne, B.M. (1998). Structural equation modelling with LISREL, PRELIS, and SIMPLIS: Basic concepts, applications, and programming. New Jersey: Lawrence Erlbaum Associates.
Devinsky, O., Souhel, N. (1999). Evidence against the existence of a temporal lobe epilepsy personality syndrome. Neurology 53, S13-25.
Meredith, W. (1993). Measurement invariance, factor analysis and factorial invariance. Psychometrika, 58, 525-543.
Tanaka, JS., & Huba, G.J. (1984). Confirmatory hierarchical factor analysis of psychological distress measures. Journal of Personality and Social Psychology, 46, 621-635.
Widaman, KF., Reise, SP. (1997). Exploring the measurement invariance of psychological instruments: Applications in the substance use domain. In Bryant, KJ. et al., (Eds).
The science of prevention: Methodological advances from alcohol and substance abuse research. (pp. 281-324). Washington, DC, American Psychological Association.
Address for correspondence: [email protected]