Is the Black Dog Is the Black Dog Really a Dalmatian?Really a Dalmatian?
An investigation into whether Stress Impact and Attributional Style lead to different
outcomes in CBT for Depression
K. Good, Massey University, New ZealandP. Merrick, Massey University, New ZealandR. Fletcher, Massey University, New ZealandN. Kazantzis, La Trobe University, Australia
o Research topic
o Methodology• Measures• Defining variables• Multilevel analysis
o Results
o Discussion
OverviewOverview
To get you thinking....To get you thinking....
o Depression and CBT• What causes variation in client recovery?
o Stress-Diathesis Model
Stress Impact
Research GapsResearch Gaps
Depression
Diathesis
Context
StressLoad
Capacity
Chronicity
Financially-related
Type of stress
Bio
Psycho
Social
Attributional Style
Existing ResearchExisting Researcho Stress and Depression
• Lack of consensus on definition (and measures)• Transition to a dynamic, contextual relationship
o Attributional Style and Depression• Depressogenic attributional style• Need more clinical samples and change over
time
o Methodological Limitationso Mostly pre-post outcomeso Few studies examining multiple predictors
simultaneously
o Patients presenting with first-time depression
o 20 Sessions of CBT• Rigorous therapist training and supervision
• Strict homework protocol and measures
o Patient measures• Beck Depression Inventory (BDI-II)
• Every session
• Attributional Style Questionnaire (ASQ)• Intake, sessions 5, 8, 20 and follow-ups
• Stress Impact ?
Treatment Study ContextTreatment Study Context
• Definition?• Objective vs Subjective
measurement?• Frequency of measurement?
Measuring StressMeasuring Stress
Stage 1 Identifying Stressors
o Therapist Questionnaire
• List top 3 stressors & rate severity
• Rate each stressor on 5 dimensions:
• relationships• autonomy• finances• control• chronicity
• Guidelines & examples developed
• Reviewed by 4 Clinical Psychologists
o Impact of Event Scale (IES)• Client questionnaire about specific stressors • 15 items (8 avoidance, 7 intrusion)
o Application to my study:• Measured at 8 time points across therapy• 0 – 5 likert scale & an optional score of “Not Rated”• Anchors (language & affect)
o 3 phases of rating:
Training Rating Checking
Measuring StressMeasuring Stress
Stage 2 Measuring Stress Across Time
Measuring Stress - Measuring Stress - ReliabilityReliability
Inter-Rater Reliability
o Percentage Agreement:o 84% agreement across all training sessionso 100% agreement at the end of the training phaseo Different sessions & therapists used in the training
Item Analysis and Reliability of the Measure
o Item AnalysisIntrusion Items Avoidance Items
1 2 3 4 5 6 71 2 3 4 5 6 7 8
3 4 5 2 4 6 7
Initial Items
Final Items
Included
How to analyse the data?How to analyse the data?Depression severityo Raw data
Stress Impact
Attributional Style
o Level 1 (time variant) vs Level 2 (time invariant)?
o Clear patterns in patient improvement by session 10
• Rapid (> 66%)• Expected (33% – 66%)• Minimal (< 33%)
o Mostly depressive vs mostly not
o Data from first three measurements
Preliminary results show Preliminary results show there is clear variability in there is clear variability in the data...the data...
0 1 3, 5, 8, 92 1, 1, 3, 3, 4, 4, 5, 6, 8, 83 0, 2, 2, 3, 3, 7, 74 2, 3, 4, 6, 95 0, 3
0 8, 91 0, 1, 3, 4, 6, 8, 92 1, 1, 2, 4, 6, 63 2, 44 2, 65
Intake BDI Scores
Rate of Change in BDI
• There is great variation among individuals• Clients vary both in their intake BDI scores
and also the rate of change they show through therapy
• Aggregating their data would neglect this variation
Building the Multilevel Building the Multilevel ModelModel
Model A
BDI
Model B
BDI Time
Model C
BDI Time Stress Curve
Model D
BDI Time Stress Curve
A-Style
```````````
• BDI vs BDI Cube Root?• Time vs Session Time?• StressCurve or Attributional Style entered first?• Stress as level 1 or level 2 (cut the data set)?
??
Parameter Model A Model B Model C Model DFixed Effects
Initial status, Intercept γ00 18.73*** 24.92***
π0i (1.82) (2.15)
Stress Curve γ01 - - - -
Attributional Style γ02 - - - - Rate of Change,
π1i Intercept γ10 - -0.93*** - (0.10)
Stress Curve γ11 - - - -
Attributional Style γ12 - - - - Variance Components
Level 1 Within-person σ2e 61.82*** 27.72*****
(3.91) (1.80)
Level 2 In initial status σ20 83.11*** 116.41***
(23.89) (33.25)
In rate of change σ21 - 0.29*
- (0.12)
Covariance σe - -4.65* - (1.86)
Pseudo R2 statistics and Goodness of Fit
R2e - 55%
R20 - -
R21 - - Deviance 3,748.63 3,391.92 AIC 3,754.63 3,403.92 BIC 3,767.42 3,429.51
Average patient had a significant change in depressive symptoms
Time explains 55% of within-patient variance
Parameter Model A Model B Model C Model DFixed Effects
Initial status, Intercept γoo 18.73*** 24.92*** 26.29***
π0i (1.82) (2.15) (4.30)
Stress Curve Leγ01 - - -1.21 - - (3.24)
Attributional Style Γ02 - - - - - - Rate of Change,
π1i Intercept γ10 - -0.93*** -.051** - (0.10) (0.18)
Stress Curve γ11 - - -0.36* - - (0.14)
Attributional Style γ12 - - - - - - Variance Components
Level 1 Within-person σ2e 61.82*** 27.72***** 27.70***
(3.91) (1.80) (1.80)
Level 2 In initial status σ20 83.11*** 116.41*** 115.48***
(23.89) (33.25) (32.99)
In rate of change σ21 - 0.29* 0.16*
- (0.12) (0.06)
Covariance σe - -4.65* -2.63* - (1.86) (1.13)
Pseudo R2 statistics and Goodness of Fit
R2e - 55% 55%
R20 - - 1%
R21 - - 45% Deviance 3,748.63 3,391.92 3,381.98 AIC 3,754.63 3,403.92 3,397.98 BIC 3,767.42 3,429.51 3,342.10
Patients with different stress curves do not have significantly different intake BDIs or rates of change
Introducing only stress as a predictor doesn’t explain a significant amount of variance in the model
Parameter Model A Model B Model C Model DFixed Effects
Initial status, Intercept γoo 18.73*** 24.92*** 26.29*** 24.47***
Π0i (1.82) (2.15) (4.30) (3.65)
Stress Curve γ01 - - -1.21 -4.75 - - (3.24) (3.87)
Attributional Style γ02 - - - 12.79** - - - (3.87) Rate of Change,
Π1i Intercept γ10 - -0.93*** -.051** -0.45* - (0.10) (0.18) (0.17)
Stress Curve γ11 - - -0.36* -0.26 - - (0.14) (0.14)
Attributional Style γ12 - - - -0.37 - - - (0.18)Variance Components
Level 1 Within-person σ2e 61.82*** 27.72***** 27.70*** 27.70***
(3.91) (1.80) (1.80) (1.80)
Level 2 In initial status σ20 83.11*** 116.41*** 115.48*** 80.23***
(23.89) (33.25) (32.99) (23.25)
In rate of change σ21 - 0.29* 0.16* 0.13
- (0.12) (0.06) (0.05)
Covariance σe - -4.65* -2.63* -1.61 - (1.86) (1.13) (0.85)Pseudo R2 statistics
and Goodness of Fit
R2e - 55% 55% 55%
R20 - - 1% 31%
R21 - - 45% 55% Deviance 3,748.63 3,391.92 3,381.98 3,372.75 AIC 3,754.63 3,403.92 3,397.98 3,392.75 BIC 3,767.42 3,429.51 3,342.10 3,435.41
After controlling for attributional style, the difference in BDI rate of change for patients with different stress curves is significantly reduced
Taken together, stress and attributional style explain 45% of variability in BDI initial status and 55% of variability in BDI rate of change
Attributional Style moderates Attributional Style moderates the the
Stress-BDI relationshipStress-BDI relationship
Rap
id
Minimal Expected Rapid
BDI-II Improvement
Stress Improvement
Exp
ecte
dM
inim
al
A depressognic attributional
style slows the improvement in BDI-II caused by an improvement
in stress
Non-depressogenicDepressogenic
Preliminary Results Preliminary Results SuggestSuggest
Attributional Style (moderator)
Stress Impact Depression
(predictor)• Stress & BDI-II correlate best when clients have
non-depressogenic AS• When they have mostly depressogenic AS, stress
improvement is stunted – it doesn’t directly correspond to improved BDI-II scores
Post-Hoc AnalysesPost-Hoc AnalysesStressor Characteristicso Modal stressor: control, relationships, autonomy,
finances, sequenceo Impact on autonomy: therapy completion and
attributional styleo Impact on relationship: previous therapy; lower stress
at conclusiono Impact on finances: therapy completiono Controllability: previous therapyo Chronicity: less severe intake BDIo Severity: gender and children
Final model variableso Stress curve, attributional style, BDI intake severity,
gender, therapy completion, marital status
Final Multilevel ModelFinal Multilevel Model
DiscussionDiscussionResearch Implicationso We can reliably access the stress people bring to
clinical situations and its effect on current functioning
o We have developed a strong hypothesis in the relationship between stress, attributional style & depression
o We can incorporate specific stressor characteristicso We have demonstrated the value of multilevel
analysis
Clinical Implicationso Improvement in stress drives improvement in
depressiono But the effect of a depressive attributional style is
powerfulo Demonstrates value of addressing attributional
style early in therapy
Selected ReferencesSelected ReferencesAbramson, L. Y., Metalsky, G. I., & Alloy, L. B. (1989). Hopelessness depression: A theory-based subtype of depression. Psychological Review, 96, 358-372.
Abramson, L. Y., Seligman, M. E. P., & Teasdale, J. D. (1978). Learned Helplessness in Humans: Critique and Reformulation. Journal of Abnormal Psychology, 87(1), 49-74.
Beck, A. T. (1976). Cognitive therapy and the emotional disorders. New York: International Universities Press.
Beck, J. (1995). Cognitive Therapy: Basics and Beyond. New York: Guilford Press.
Beck, A. T., Rush, A., Shaw, B., & Emery, G. (1979). Cognitive therapy of depression. New York: The Guilford Press.
Cohen, S., Kessler, R. C., & Underwood Gordon, L. (1997). Measuring Stress: A Guide for Health and Social Scientists. New York, USA: Oxford University Press.
Selected ReferencesSelected ReferencesHammen, C. (2005). Stress and Depression. Annual Review of Clinical Psychology, 1, 293-319.
Hedeker, R., & Gibbons, R. D. (2006). Longitudinal Data Analysis. New Jersey, USA: John Wiley & Sons, Inc.
Horowitz, M., Wilner, N., & Alvarez, W. (1979). Impact of Event Scale: A Measure of Subjective Stress. Psychosomatic Medicine, 41(3), 209-218.
Kaplan, H. B. (1996). Psychosocial Stress: Perspectives on Structure, Theory, Life-Course, and Methods. London: Academic Press, Inc.
Lambert, M. J. (2004). Bergin And Garfield’s Handbook of Psychotherapy and Behavior Change. New York, USA: John Wiley & Sons, Inc.
Singer, J., & Willett, J.B. (2003). Applied Longitudinal Analysis: Modeling Change and Event Occurrence. New York: Oxford University Press.
20090312 Doctoral Confirmation Presentation.ppt
Thank you!Thank you!
Contact:
Kimberly GoodDClinPsych candidateMassey University (Auckland, New Zealand)Email: [email protected]