direction of causation modeling between measures of distress and parental bonding
DESCRIPTION
Direction of Causation Modeling Between Measures of Distress and Parental Bonding. N.A. GILLESPIE 1 , D. DUFFY 1 , G. ZHU 1 , A.C. HEATH 2 , N.G. MARTIN 1. 1 Queensland Institute of Medical Research and University of Queensland, Joint Genetics Program, Brisbane, Australia. - PowerPoint PPT PresentationTRANSCRIPT
Direction of Causation Modeling Between Measures of Distress and
Parental Bonding
N.A. GILLESPIE1, D. DUFFY1, G. ZHU1, A.C. HEATH2, N.G. MARTIN1
1 Queensland Institute of Medical Research and University of Queensland, Joint Genetics Program, Brisbane, Australia.
2 Department of Psychiatry, Washington University School of Medicine, St Louis, Missouri, USA
Supported by NIH grants AA04535, AA07728, and AA10249 and NHMRC (Australia) grants 941177and 971232.
Relationship between Parental Bonding and Psychiatric Symptoms
• Parental Bonding Instrument (PBI) (Parker) - Maternal and Paternal Care and Overprotection- Low Care and High Overprotection Depression and Anxiety - Replicates across populations
• PBI stable over time
- During and after depressive episodes- Not artifact of previous depression episodes- Children’s and Parent’s self-ratings (Care r=0.44, Overprotection r=0.55)
Causes of association between variablesA B Unidirectional causationA B Unidirectional causationA B Reciprocal causation A C Indirect causation (common A and / or C effects) B
Instrumental variable methods- Restrictive assumptions- Unfeasible for psychiatric epidemiology and behavioural sciences
Panel / Longitudinal Data- Costly- Underlying assumptions and errors of inference
Random Assignment to experimental and control Groups- Unethical and unworkable
Twin DataAvoids difficulties mentioned & cost effectiveCausal inferences possible using cross-sectional data
MZ=(a2+d2)c2iBDZ=(½a2+¼d2)c2iB
aBdB aB
1
1MZ, ½DZ
1MZ, ¼DZ
dBeB eB
cA eAeA cA
iB iB
At1
Bt1
E A D
E C
At2
Bt2
D A E
C E
aBdB aB
1
1MZ, ½DZ
1MZ, ¼DZ
dBeB eB
cA eAeA cA
iA iA
At1
Bt1
E A D
E C
At2
Bt2
D A E
C E
MZ=c2iBDZ=c2iB
A B B A
Previous findings
• Depression (CESD) and Parental Bonding Instrument (PBI) - Unidirectional Model: Parental Behaviour
Depression • Results equivocal with inclusion of error measures• Common additive genetic factor model provided the best fit
Limitations of DOC models
• Failure to included measurement error leads to biased estimates - Reduced statistical power to resolve alternate causal
hypotheses - Unless can estimate error (test-retest) or assume it is negligible
• To test DOC hypotheses between Parental Bonding and psychiatric symptoms using multiple measures of the target variables– Multiple Indicator Model (Heath et al. 1992) – Inclusion of test-retest data or error terms not required– Assumes that measurement error occurs at the level of indicator
variables for the PBI and psychiatric symptom measure
Objective
MethodSample
• Combined community-based sample of:- 3041 Female twin individuals (25-45yrs) Alcohol Cohort Study 1- 4397 Female twin individuals (18-28yrs) Alcohol Cohort Study 2
Psychiatric symptom measures
• Both cohorts received a 33 item self-report symptom inventory - 19 items from Symptom Checklist-90 (Derogatis et al.,1973) - 14 items from Delusion Symptoms States Inventory (Foulds & Bedford, 1975)
• Recently experienced psychological distress:“Recently have blamed myself for things”“Recently have experienced pain or tension in neck/head”
• (1) ‘not-at-all’ (2) ‘a little’ (3) ‘a lot’ (4) ‘unbearably’
Parenting Measures
• 7 item self-report Parental Bonding Instrument (PBI) - 2 Care items “Made me feel I was not wanted”- 5 Overprotection items “Tried to make me dependent on him/her”
• “When you were growing up, how much were your mother and father (or someone like mother and father) like this?”
• (1) ‘not-at-all’ (2) ‘ a little’ (3) ‘somewhat’ (4) ‘a lot’
Factor analysis of psychiatric symptoms
Based on inter-item polychoric correlations with Promax rotation: - Depression- Anxiety- Somatic Distress- Sleep Disturbance
Factor analysis of PBI items
Based on inter-item polychoric correlations with Promax rotation: - Autonomy - Overprotection- Coldness
Authoritarianism, Protectiveness, and Warmth (Kendler)
Twin correlations
MZ DZtwin pairs=969 twin
pairs=608.37 .14 .43 .26.34 .19
.50 .36
.47 .34
.61 .35
Reliability
r
Depression .60 .86Anxiety .56 .83Somatization Distress .71 .75
N = 420
Parental Autonomy .65 .80Parental Overprotection .70 .77Parental Coldness .60 .75
N = 393
Univariate AnalysesVariance components WLS Estimates of best fitting models
Variable A C E D AGE 2 df
Overprotection .27 .20 .53 .00 1.06 5 Coldness .62 .37 .01 6.99 6 Autonomy .31 .18 .50 .01 8.99 5
Depression .31 .65 .04 3.50 6 Anxiety .41 .56 .03 1.62 6 Somatic Distress .34 .65 .01 4.85 6
Independent-Pathway model for ratings of Parental Coldness, Autonomy and Overprotection
COLD OVERP AUTO
EA EA E A
A EC
CC C
Model 2 df p AIC- Independent Pathway 9.17 15 .87 -20.83
Model 2 df p AIC- Independent Pathway 9.17 15 .87 -20.83- Common Pathway 13.39 19 .82 -24.61
COLD OVERP AUTO
EA EA E A
A EC
CC C
PARENTING
Common-Pathway model for ratings of Parental Coldness, Autonomy and Overprotection
.64.12.44
.24.27.49
.16.12
.28 .21.17
.49 . 00.03
.33
Independent Pathway model for ratings of ratings of ratings of Depression, Anxiety, and Somatic Distress
Model 2 df p AIC- Independent Pathway 19.50 15 .19 -10.50
DEP ANX SOM
EA EA E A
A EC
CC C
DISTRESS
DEP ANX SOM
EA EA E A
Model 2 df p AIC- Independent Pathway 19.50 15 .19 -10.50- Common Pathway 24.06 19 .19 -13.94
CC C
A EC
Common Pathway model for ratings of ratings of ratings of Depression, Anxiety, and Somatic Distress
.42 .52
.61.67
.48
.31.08 .22.12 .42.11 .00.00.00
.06
Model 2 df 2 AIC Full Bivariate 145.66 107 -69.34
COLD
C EA
AUTON
C EA
OVERP
C EA
A EC
DEP
C EA
SOM
C EA
ANX
C EA
DISTRESS PARENTING
A EC
Full Bivariate Model with Multiple Indicators
Reciprocal Causation Model with Multiple Indicators
Model 2 df 2 AIC Full Bivariate 145.66 107 -69.34Reciprocal 146.00 108 .34 -70.00
COLD
C EA
AUTON
C EA
OVERP
C EA
A EC
DEP
C EA
SOM
C EA
ANX
C EA
DISTRESS PARENTING
A EC
Uni-directional Model with Multiple Indicators
Model 2 df 2 AIC Full Bivariate 145.66 107 -69.34Reciprocal 146.00 108 .34 -70.00Distress Parenting 161.74 109 16.08 -56.26
COLD
C EA
AUTON
C EA
OVERP
C EA
A EC
DEP
C EA
SOM
C EA
ANX
C EA
DISTRESS PARENTING
A EC
Uni-directional Model with Multiple Indicators
Model 2 df 2 AIC Full Bivariate 145.66 107 -69.34Reciprocal 146.00 108 .34 -70.00Distress Parenting 161.74 109 16.08 -56.26Parenting Distress 146.71 109 1.05 -71.29
COLD
C EA
AUTON
C EA
OVERP
C EA
A EC
DEP
C EA
SOM
C EA
ANX
C EA
DISTRESS PARENTING
A EC
Model 2 df 2 AIC Full Bivariate 145.66 107 -69.34Reciprocal 146.00 108 .34 -70.00Distress Parenting 161.74 109 16.08 -56.26Parenting Distress 146.71 109 1.05 -71.29No causation 376.29 110 230.63 156.29
COLD
C EA
AUTON
C EA
OVERP
C EA
A EC
DEP
C EA
SOM
C EA
ANX
C EA
DISTRESS PARENTING
A EC
No Causation Model with Multiple Indicators
C E C EC EAC E EAEA
Model 2 df 2 AIC Full Bivariate 145.66 107 -69.34Reciprocal 146.00 108 .34 -70.00Distress Parenting 161.74 109 16.08 -56.26Parenting Distress 146.71 109 1.05 -71.29No causation 376.29 110 230.63 156.29Final 151.26 116 5.60 -80.74
Final Uni-directional Model with Multiple Indicators
COLD AUTONOVERP
A E
DEP SOMANX
DISTRESS PARENTING
A EC
.38 .45
.63 .67 .49 .56.16 .52
.55.20 .25
.17 .26 .21 .14 .49 .11 .37.36 .13 .21 .11 .40
+ .18
Conclusion
- Almost 50% of variance in Distress attributable to additive genetic effects
- Considerable proportion of variance in Parenting due to common environmental effects
- Recollections of Parenting appear to have a causative effect of symptoms of distress
Limitations- Extrapolation of findings to males
- Retrospective reports of parental style
- Kendler- Parental self-ratings of parental style large due to C- Twins ratings of parental style largely due to A, C, E- Correlations between parents and offspring 0.10 - 0.29
- Correlations between twins 0.37 - 0.63
- Results can only be considered a ‘partial reflection’ of true parenting
- PBI is not exhaustive measure