causal attributions in clinical subtypes of depression: a longitudinal study of inpatients

Journal of Psychopathology and Behavioral Assessment, 1/ol. 13, No. 3, 1991

Causal Attributions in Clinical Subtypes of Depression: A Longitudinal Study of Inpatients

Asle Hoffart 1,2 and Egil W. Martinsen 1 Accepted: September 17, 1991

The study material comprised inpatients who met DSM-III-R criteria for (a) dysthymia without panic and~or agoraphobia (n = 20), (b) major depression without panic and~or agoraphobia (n = 26), (c) both major depression and panic with agoraphobia (comorbid patients) (n = 17), and (d) panic with agoraphobia without any depressive disorder (n = 22). The patients completed the Attributional Style Questionnaire and the Beck Depression Inventory and were assessed on the Comprehensive Psychopathological Rating Scale upon admission to the hospital and at discharge. Some of the self-report scales were also administered at 1-year follow-up. It was assumed that dysthymic patients and patients with both major depression and agoraphobia wouM exhibit more biased attributions for bad events than "purely" major depressed and "purely" agoraphobic patients. However, inconsistent with this hypothesis, obtained group differences could be statistically reduced to differences in depressive symptom level At each assessment, attributions for bad events correlated significantly with depressive symptom level Attributional bias tended to decrease during treatment. However, most attribution subscales exhibited moderate stability in terms of correlation across assessments. Attributing bad events to global causes proved to predict later depression.

KEY WORDS: attributions; depressive subtypes; longitudinal.

This research was supported by grants from the Halldis and Josef Andresen Legacy. 1Research Institute, Modum Bads Nervesanatorium, N-3370 Vikersund, Norway. 2To whom correspondence should be addressed.

241

242 lloffart and Martinsen

INTRODUCTION

We all have a tendency to attribute causes to events that are important to us. In the hopelessness theory of depression, Abramson, Alloy, and Metalsky (1989) postulated that the habit of attributing bad events to stable and global causes and assigning importance to these events represents a vulnerability factor for the occurrence of depression. Thus, this model hy- pothesizes that attributions predict later depression. Earlier studies also indicated that this is the case (Brewin, 1985). However, no studies have investigated the unique predictive power of each of the relevant attributional dimensions. The hopelessness model implies that the stability, globality, and importance dimensions all contribute independently to later depression.

Further, the hopelessness model presupposes that causal attributions express rather enduring personality traits. Another view is that cognitive biases are correlates or consequences of the depressive state itself. Whereas the hopelessness model predicts that attributional biases will persist beyond the remission of a depressive episode, the latter view implies that the biases will dissipate when symptoms remit. The evidence pertaining to this question is inconclusive. Hamilton and Abramson (1983) obtained differences in attributions for hypothetical events on the Attributional Style Question- naire (ASQ; Peterson et aL, 1982) between depressed and nondepressed patients. Upon remission, however, these differences disappeared. Con- trasting with these results, Eaves and Rush (1984) reported "depressogenic" attributions in depressed patients that were different from the control group and that persisted after the major symptom had disappeared. A possible explanation of these divergent findings is that both the hopelessness model and the symptom perspective are partially valid. Abramson et al. (1989) suggested that the state of depression may strengthen already ex- isting traitlike cognitive biases. If this is true, attributional biases should decline with remission, but the rank order between individuals on the attributional dimensions should be fairly stable from a symptomatic to a remitted state.

The hopelessness model is supposed to explain only a specific type of depression (Abramson et aL, 1989). So far, however, this subtype has not been clearly identified. Hamilton and Abramson (1983) raised the pos- sibility that individuals suffering from chronic depressive syndromes, like dysthymia, may exhibit traitlike, depressogenic causal attributions. A stable tendency to think negatively should logically lead to some depressive symptoms most of the time, as the individual would be expected to experience at least minor negative events regularly. Alloy, Kelly, Mineka, and Clements (1990) suggested that many hopelessness depressive patients may also be

Causal Attributions in Depressive Subtypes 243

suffering from anxiety. This follows from the logic of the model: Hopeless- ness is a sufficient "proximal" cause of depression. Hopelessness includes both negative expectations about the occurrence of highly valued outcomes (a negative outcome expectancy) and feelings of helplessness about chang- ing the likelihood of occurrence of these outcomes. However, a sense of helplessness in controlling important future outcomes normally leads to anxiety. Individuals who suffer from depression that is based on hopelessness should therefore exhibit anxiety symptoms as well. Alloy et al. (1990) also suggest that several features of anxiety-depression comorbidity may be predicted from the hopelessness theory, for instance, that anxiety without depressive symptoms is more common than depression without anxiety symptoms. Recent cross-sectional studies have linked causal attributions to anxiety disorders and depression. Heimberg et al. (1989) found that dysthymic patients did not differ from agoraphobic and social phobic patients regarding attributions. Hoffart and Martinsen (1990a) compared depressed and agoraphobic patients and failed to find the differences predicted from hopelessness theory.

The present longitudinal study examined causal attributions in dysthymic patients, major depressed patients, patients with both agoraphobia and major depression, and nondepressed agoraphobic patients. The study represents an extension of the previous cross-sectional study (Hoffart & Martinsen, 1990a), but the depressed patients were now divided into the mentioned diagnostic subtypes. Patients with agoraphobia and dysthymia were excluded from the group comparison part of this study because they were too few to form a group of their own. Furthermore, only patients who completed the repeated assessments were included. According to the rationales given above on possible "hopelessness" subtypes of depression, our first aim was to examine whether attributional biases were more promi- nent among dysthymic and comorbid patients. We therefore predicted the following. (1) Dysthymic and comorbid patients will attribute bad events to more stable and global causes, and assign more importance to these events, than patients with major depression or agoraphobia alone. These differences will persist over time and not be reducible to differences in depressive symptom level. To test the assumption that symptoms will pro- duce feedback on attributions, we set up two further predictions. (2a) After treatment, when depressive symptoms have presumably abated, the patients will display smaller biases on the bad events stability, globality, and importance dimensions compared to themselves when they were acutely depressed. (2b) Bad events stability, globality, and importance scores will correlate with depressive symptom level. Finally, we wished to test the two presuppositions of the hopelessness model addressed above. (3) Attribution scores will exhibit stability over time in terms of correlations across assess-

244 Hoffart and Martinsen

ments. (4) Bad events stability, globality, and importance scores will contribute independently to later depression.

METHOD

Subjects

The sample consisted of inpatients at a Norwegian psychiatric insti- tution. The treatment program includes psychodynamically oriented psychotherapy, occupational and milieu therapy, physical exercise, and, in some cases, pharmacotherapy. A specialized exposure-based agoraphobia treatment program has been developed (Hoffart & Martinsen, 1990b). Con- secutive patients were administered the SCID interview (Spitzer & Wil- liams, 1984) for DSM-III-R Axis I and II diagnoses (American Psychiatric Association, 1987) shortly after admission. The group comparison part of this study covered patients who met the criteria for (a) dysthymia without panic and/or agoraphobia (n = 20), (b) major depression without panic and/or agoraphobia (n = 26), (c) both agoraphobia and major depression (comorbid patients; n = 17), and (d) agoraphobia without any depressive disorder (n = 22). Patients who met criteria for panic disorder and/or agoraphobia were not further questioned about other anxiety disorders, and patients who met criteria for major depression were not asked about dysthymia. Thus, some of the major depressed patients might have been dysthymic as well. The second, predictive part of this study additionally included nine patients who met criteria both for panic disorder with agoraphobia and dysthymia. The SCID interviews were conducted by one of the authors (E.W.M.). To assess the diagnostic reliability, 21 randomly selected audiotaped SCID interviews were independently scored by another psychiatrist. Interrater reliability was calculated by the use of kappa. The kappa value was .84 on Axis I and .76 on Axis II. Kappa values for individual Axis I diagnoses were .84 for dysthymia and agoraphobia and .90 for major depression.

Instruments and Procedure

Causal attributions were measured by the authors' Norwegian trans- lation of the ASQ. Here subjects were asked to imagine experiencing 12 events, 6 good and 6 bad, and to determine the possible major cause of each event. Next, they rated these causes on 7-point scales of internality, stability, and globality. The actual event was also rated for importance. In-


ternal consistencies for good events were estimated, using Cronbach's alpha, to .67, .57, .61, and .56 for the internality, stability, globality, and importance subscales, respectively. For bad events, the corresponding values were .69, .76, .62, and .71. On average, these alpha values were higher than those found in the original studies of the ASQ (Peterson et al., 1982). Factor analyses with varimax rotation of the ratings for internality, stability, globality, and importance on the 12 events supported a two-factor solution for all four dimensions, one for good and one for bad events. In each case, a scree test indicated two factors, and almost every item loaded (>.40) exclusively on its expected factor. This is consistent with the original findings (Peterson et al., 1982).

The Beck Depression Inventory (BDI; Beck, Steer, & Garbin, 1988) is a widely used and well-validated self-report instrument. The Compre- hensive Psychopathological Rating Scale (CPRS; Aasberg, Montgomery, Perris, Schalling, & Sedvall, 1978) is composed of 65 0-30 scaled items covering a wide range of psychopathology. Symptoms are rated by an in- terviewer. In our study, we rated only the 38 items that were considered relevant for non-psychotic patients. In a former study of a larger sample (N = 146), which included the patients of the present study, we identified a 12-item index of depression (CPRS-D) and a 7-item index of anxiety (CPRS-A; Martinsen, Friis, & Hoffart, 1989). The alpha values were .80 and .82 for the anxiety and depression indices, respectively. The indices differentiated well between patients with anxiety and those with depressive disorders among subgroups of anxiety and depression. Ratings on the CPRS were performed by either the first author or the mentioned psychiatrist. To assess interrater reliability, 20 randomly selected patients were in- terviewed and rated by the second author in the presence of the other two, who also rated these patients. The reliability was high for both indices, the intraclass correlations being .95 or above.

The patients completed questionnaires and were rated on the CPRS shortly after the diagnostic interview ("pretest") and at discharge ("posttest"). Follow-up investigation was performed one year after discharge by mailed questionnaires. Only the BDI and the internality scales of the ASQ were administered at follow-up.

Statistics

Pretherapy variables were tested for differences by F test for continu- ous variables and Z2 test for categorical variables. If the four-way group comparison revealed a difference significant at the .05 level, follow-up pair- wise comparisons were made using Duncan's test or 1-dr Z2 test. Repeated-

246 Hoffart and Mar t insen

measures ANOVA was performed with Group as a between-subjects factor and Time as a within-subjects factor. Whenever this F test yielded a significant (p < .05) group effect but no interaction, Duncan's post hoc test (p < .05) was used. Whenever the repeated-measures ANOVA revealed a significant Group x Time interaction, we performed one-way ANOVAs at each assessment to detect the source of the interaction. Hierarchical regression was applied in the predictive analyses.

RESULTS

Group Comparisons

Demography, history of illness data, and treatment data for the 20 dysthymic, 26 major depressed, 17 comorbid (major depressed and agoraphobic), and 22 agoraphobic patients are presented in Table I. ANOVAs with Duncan's post hoc tests (p < .05) indicated that the agoraphobic patients were younger than the major depressed patients. ~2 test with 1-df Z 2 follow-up test indicated that more of the agoraphobic and comorbid than of the major depressed patients came from lower social classes (skilled or unskilled worker, unemployed, insured). Personality disorders were more frequent among comorbid than among major depressed patients. In addition to "treatment as usual," many agoraphobic patients participated in the specialized exposure-based treatment program for agoraphobia. A greater number of the purely agoraphobic patients (19/22) than of the comorbid

Table I. Proportions/Means on Background and Treatment Variables for Diagnostic Groups a

Group

Variable D M MA A Test

Female 11/20 14/26 14/17 14/22 Z2 = 4.59 Lower social class 4/20 2/26 8/17 10/22 Z- = 12.73"* Age 41.3 45.4 41.8 36.9 F(3,81) = 3.05* Length of depressive episode (months) 50.2 25.2 43.6 - - F(2,59) = 1.87 Length of anxiety (years) - - - - 11.4 10.4 F2(1,38 ) = 0.45 Personality disorder 14/20 13/26 15/17 15/22 Z2 = 7.80* Exposure t reatment - - - - 7/17 19/22 Z = 9.02* Using antidepressants 3/20 14/26 8/17 4/22 Z 2 = 10.84" Length of stay (weeks) 13.8 15.7 11.2 12.6 F(3,81) = 2.29

aD, dysthymia; M, major depression; MA, major depression and phobia.

*p < .05. **p < .01.

agoraphobia; A, agora-


patients (7/17) took part in this program. More of the major depressed and the comorbid than of the dysthymic and the agoraphobic patients received antidepressants. No group differences emerged in terms of sex, length of illness, and length of stay at the hospital (overall mean = 13.4 weeks).

The symptom measures were analyzed to obtain a basis for evaluat- ing attributional differences and changes. For the BDI, repeated-measures ANOVA indicated group [F(3,81) = 3.46, p < .05], time [F(2,162) = 47.56, p < .0001], and interaction [F(6,162) = 3.71, p < .01] effects. Like- wise, for the CPRS-D index, ANOVA yielded group [F(3,81) = 5.35, p < .01], time [F(1,81) = 88.68,p < .0001], and interaction [F(3,81) = 7.14,

p < .01] effects. CPRS-D level decreased from pre- to posttest. One-way ANOVAs performed at each assessment showed that the major depressed and the comorbid patients tended to have a significantly higher BDI and CPRS-D level than the dysthymic and the agoraphobic patients at pretest (see Table II). At posttest, the comorbid group had a higher BDI-level than the other groups. For the CPRS-A index, ANOVA revealed group [F(3,81) = 19.61,p < .0001] and time [F(1,81) = 66.21, p < .0001] effects but no interaction. As expected, Duncan's post hoc test showed that the agoraphobic and the comorbid patients scored higher on the CPRS-A than the dysthymic and the major depressed patients. CPRS-A level decreased from pre- to posttest.

Repeated-measures ANOVA revealed significant group effects for the bad events stability, globality, and importance measures (see Table III). As predicted from the hopelessness model, the comorbid patients scored higher than the agoraphobic patients on all three dimensions: stability, globality, and importance. However, they scored higher than the major depressed patients only on the importance dimension. Disagreeing with our hypothesis, the dysthymic patients did not score higher than the patients in the other groups on any of the three relevant dimensions. Time effects were revealed for the bad events stability [F(1,81) = 5.21, p < .05], globality [F(1,81) -- 22.01, p < .0001], and importance [F(1,81) = 3.69, p < .05] measures. As predicted, the scores decreased from pre- to posttest on all three measures. Time effects were also noted for the internality measures, both for good events [F(2,162) = 5.58,p < .01] and for bad events [F(2,162) = 9.04, p < .001]. An interaction effect was revealed for the measure of the importance of good events [F(3, 81) = 3.40,p < .05]. However, one-way ANOVAs performed at each assessment did not reveal any group difference regarding importance ratings (see Table III).

To examine whether group differences regarding age, social class, and personality disorders did confound the results, these variables were included as constant covariates in separate repeated-measures ANCOVAs

g

Tab

le I

I, M

ean

Sym

ptom

Sco

res

for

Gro

ups

at E

ach

Tim

epoi

nt a

Gro

up

Maj

or

Maj

or d

epre

ssio

n D

ysth

ymia

de

pres

sion

an

d ag

orap

hobi

a A

gora

phob

ia

(n =

20)

(n

=

26)

(n =

17

) (n

=

22)

Mea

sure

M

(S

D)

M

(SD

) M

(S

D)

M

(SD

) F(

3,81

) b

Gro

up d

iffe

renc

es c

BD

I Pret

est

20.9

(5

.5)

27.2

(9

.0)

25.8

(9

.3)

17.5

(8

.1)

8.93

**

(M,

MA

) >

(D,

A)

Post

test

9.

6 (7

.9)

11.2

(8

.8)

18.3

(9

.4)

11.9

(8

.2)

3.38

* M

A >

(D

, M

, A

) Fo

llow

-up

9.7

(10.

9)

11.1

(1

0.0)

18

.4

(12.

6)

16.0

(1

3.1)

1.

63

CPR

S-D

Pr

etes

t 9.

0 (2

.8)

13.0

(5

.1)

11.5

(4

.0)

7.4

(3.9

) 8.

75**

M

>

(D,

A);

MA

> A

Po

stte

st

4.2

(3.7

) 4.

9 (4

.3)

7.2

(5.1

) 5.

1 (3

.6)

1.63

C

PRS-

A

Pret

est

9.7

(5.2

) 8.

8 (4

.8)

17.0

(4

.5)

15.9

(3

.2)

Post

test

6.

3 (5

.1)

5.2

(4.4

) 13

.2

(4.6

) 11

.9

(4.6

) 19

.61"

**

(MA

, A

) >

(D,

M)

aBD

I, B

eck

Dep

ress

ion

Inve

ntor

y; C

PRS,

Com

preh

ensi

ve P

sych

opat

holo

gica

l R

atin

g Sc

ale;

-D

, D

epre

ssio

n In

dex;

-A

, A

nxie

ty I

ndex

. D

, dy

sthy

mia

; M

, m

ajor

dep

ress

ion;

MA

, m

ajor

dep

ress

ion

and

agor

apho

bia;

A,

agor

apho

bia.

bG

roup

res

ults

(po

oled

ove

r ti

me)

of

Gro

up •

T

ime

AN

OV

As

whe

n no

int

erac

tion

occ

urre

d. W

hen

the

Gro

up •

Tim

e A

NO

VA

gav

e a

sign

ific

ant

inte

ract

ion,

the

res

ults

of

one-

way

AN

OV

As

perf

orm

ed a

t ea

ch t

imep

oint

are

rep

orte

d.

CR

esul

ts o

f D

unca

n's

post

hoc

tes

ts (

p <

.05)

. *p

<

.05.

**

p <

.001

. **

*p <

.0

001.

O

Tab

le I

Il,

Mea

n A

ttri

buti

on S

core

s fo

r G

roup

~ at

Eac

h T

imep

oint

,n

.

Dys

thym

ia

(n =

20)

Gro

up

Mea

sure

M

(S

D)

Maj

or

Maj

or d

epre

ssio

n de

pres

sion

an

d ag

orap

hobi

a A

gora

phob

ia

(n =

26

) (n

=

17)

(n =

22

)

M

(SD

) M

(S

D)

M

(SD

) F(

3,81

) b

Gro

up d

iffe

renc

es c

O =.

Goo

d ev

ents

In

tern

alit

y Pr

etes

t 4.

20

(1,1

4)

3,97

(1

.22)

3.

04

(0.9

2)

4.33

(1

.40)

Po

slte

sl

4.51

(0

.72)

4.

43

(1,3

2)

3,69

(]

.33)

4.

54

(1,2

0)

Fotlo

w-u

p 4.

19

(0.8

6)

4,32

(1

,14)

3.

95

(1.3

8)

4,47

(0

.84)

St

abili

ty

Pret

est

4.73

(0

.85)

4.

83

(0.8

2)

5.05

(i

,20)

5.

02

(0.8

4)

Post

test

4.

74

(0.8

0)

4.64

(0

.85)

4.

74

(1.2

4)

5.07

(0

.77)

G

loba

lity

Pr

etes

t 4.

43

(1.2

4)

4.37

(1

.08)

5.

00

(1.3

3)

4.59

(1

.13)

Po

stte

st

4.02

(1

.01)

4.

50

(0.6

9)

4.99

(1

.31)

4.

27

(0.7

4)

Impo

rtan

ce

Pret

est

5.67

(0

.87)

5,

54

(0.7

8)

5.40

(0

.94)

5.

61

(0.7

2)

Post

test

5.

26

(1.0

2)

4.50

(0

.86)

5.

81

(0.8

6)

5.78

(0

.82)

B

ad e

vent

s In

tern

alit

y Pr

etes

t 5.

13

(0.9

3)

5,39

(1

.08)

5.

14

(1,3

9)

5.33

(1

.02)

Po

stte

st

4.60

(0

.83)

4.

94

(1,0

4)

5.41

(1

.10)

5,

26

(0.8

4)

Foll

ow-u

p 4.

36

(1.1

6)

4.85

(]

,10)

4.

50

(1.3

7)

4.98

(0

,94)

2.80

*

1.03

2.46

0.23

1.

15

1.04

(D,M

,A)

>M

A

It~

Tab

le I

II.

Con

tinu

ed

Gro

up

Maj

or

Maj

or d

epre

ssio

n D

ysth

ymia

de

pres

sion

an

d ag

orap

hobi

a A

gora

phob

ia

(n =

20)

(n

= 2

6)

(n =

17

) (n

= 2

2)

Mea

sure

M

(S

D)

M

(SD

) M

(S

D)

M

(SD

) F(

3,81

) b

Gro

up d

iffe

renc

es c

Stab

ilit

y Pr

etes

t 4.

39

(1.2

3)

5.28

(0

.87)

5.

33

(0.9

9)

4.31

(1

.06)

Po

stte

st

4.27

(0

.87)

45

3 (0

.95)

5.

30

(1.4

0)

4.16

(0

.90)

6.

42**

(M

, M

A)

> (D

, A

) G

loba

lity

Pr

etes

t 4.

53

(1.1

3)

4.95

(0

.91)

5.

31

(1.0

8)

4.31

(1

.05)

Po

stte

st

3.96

(0

.87)

4.

19

(0.9

5)

4.65

(1

.56)

3.

88

(1.1

7)

3.55

* M

A >

(D

, A

) Im

port

ance

Pr

etes

t 5.

58

(1.0

0)

6.01

(0

.73)

5.

89

(0.9

6)

5.43

(1

.00)

Po

stte

st

4.98

(1

.08)

5.

58

(0.9

7)

6.19

(0

.76)

5.

26

(1.3

6)

3.31

" M

A >

(D

, M

, A

)

a D

, dy

sthy

mia

; M

, m

ajor

dep

ress

ion;

MA

, m

ajor

dep

ress

ion

and

agor

apho

bia;

A,

agor

apho

bia.

b

Gro

up r

esul

ts o

f G

roup

• T

ime

AN

OV

As

whe

n no

int

erac

tion

occ

urre

d. W

hen

the

Gro

up x

Tim

e A

NO

VA

gav

e a

sign

ific

ant i

nter

acti

on,

the

resu

lts

of o

ne-w

ay A

NO

VA

s pe

rfor

med

at

each

tim

epoi

nt a

re r

epor

ted.

C

Res

ults

of

Dun

can'

s po

st h

oc t

ests

(,o

<

.05)

. *p

<

.05.

**

p <

.001

. o .z.

~

t


on the attributional measures. Correcting for age and social class made no change in the results. After controlling for the presence of personality disorder, the group effect on the good events internality measures disappeared. The group effects remained on the bad events stability [F(3,80) = 5.90, p < .001], globality [F(3,80) = 3.05, p < .05], and importance [F(3,80) = 3.08, p < .05] measures. However, post hoc tests revealed some changes in that comorbid patients no longer scored higher than the dysthymic patients on the globality measure and no longer scored higher than the major depressed patients on the importance measure. The potential effect of the additional exposure program was examined by comparing those agoraphobic and comorbid patients who participated with those who did not. AN- COVAs on the posttest attribution scores with pretest scores as covariate yielded no significant results (all F's < 1). Because diagnosis and exposure treatment could interact, we repeated these ANCOVAs including only the comorbid patients. There were no significant differences between the 7 comorbid patients who participated and the 10 who did not (all F's < 1). Too few purely agoraphobic patients were nonparticipants in the exposure program to make meaningful comparisons within the agoraphobic group.

In another set of analyses, repeated-measures ANCOVAs with the BDI scores at different time points as covariates were performed on the ASQ measures. All group and time effects disappeared.

To test the hypotheses concerning correlations between attributions and depressive symptom level, partial correlation analyses were performed at each assessment between ASQ scores and CPRS-D scores, with CPRS-A scores as the third variable. Of the 16 tests performed, 10 gave significant results. At pretest, the partial coefficients were significant between depression scores and good events internality [r(83) = -.32, p < .001], and importance [r(83) = -.25, p < .05] scores and bad events internality [r(83) = .29, p < .01], stability [r(83) = .40, p < .0001], globality [r(83) = .54, p < .0001], and importance [r(83) = .25, p < .05] scores. At posttest, the partials were significant between depression scores and good events internality scores [r(83) = -.23, p < .05] and bad events internality [r(83) = .42, p < .0001], stability [r(83) = .38, p < .0001], and globality [r(83) = .54, p < .0001] scores. Thus, our hypotheses regarding the bad events stability, globality, and importance dimensions were supported in almost every instance. When CPRS-D scores were partialed out of the correlations between ASQ and CPRS-A scores, only two significant coefficients occurred: for pretest bad events globality scores [r(83) = .21, p < .05] and for posttest good events stability scores [r(83) = .22, p < .05].

The partial correlations analyses were repeated within each diagnostic group. Within the dysthymic group, pre- and posttest bad events globality scores correlated significantly with CPRS-D scores [r(19) = .55, p < .05,


and r(19) = .05, respectively]. Within the major depressed group, pre- and posttest bad events internality scores correlated with CPRS-D scores [r(25) = .48, p < .05, and r(25) = .54, p < .01]. In addition, posttest bad events stability and globality scores correlated with CPRS-D scores [r(25) = .48, p < .05, and r(25) = .63, p < .01]. Within the comorbid group, pre- and posttest globality scores correlated with CPRS-D scores [r(16) = .49, p < .05, and r(16) = .64, p < .01]. In addition, posttest bad events internality and stability correlated with CPRS-D Jr(16) = .71, p < .01, and r(16) = .53, p < .05]. Within the agoraphobic group, pretest bad events globality scores correlated with CPRS-D scores [r(21) = .64, p < .01]. No significant correlations occurred at posttest. Thus, the within-group analyses revealed similar patterns across groups at pretest, whereas depressive symptoms and attributions for bad events seemed to be more closely related in the major depressed and in the comorbid group at posttest.

Predictive Analyses

In the predictive part of the study, nine additional patients were included (see Subjects). The CPRS-A scores showed high stability across pre- and posttest [r(93) = .72, p < .0001], whereas the depression scores did not [CPRS-D, r(93) = .26, p < .05; BDI, r(93) = .38, p < .001]. Overall, the ASQ subscales exhibited moderate stability. For good events, the r(93)'s across pre- and posttest were .57 (p < .0001), .12 (ns), .45 (p < .0001), and .49 (p < .0001) for the internality, stability, globality, and importance subscales, respectively. The corresponding values for bad events were .48, .54, .45, and .46 (all p's < .0001). From posttest to 1-year follow-up, the r(93)'s were .66 and .43 (both p's < .0001) for the good and bad events internality scales, respectively.

In the following regression analyses, initial level of depression was forced to enter first to partial out the explained variance due to the concurrent relation between attributions and depression. In the complete sample, pretest BDI scores predicted a significant portion of the variance in posttest BDI scores [R 2 change = .160, t(93) = 4.1, p < .0001]. Of the pretest attributional scores, only the bad events globality scores added a significant amount in explained variance [R 2 change = .077, t(92) = 3.0, p < .01]. That is, when the globality variable was included in the equation, no other variable added a significant amount. To examine the consistency of these results across groups, these analyses were repeated for the bad events globality variable within each diagnostic group. Pretest bad events globality exhibited a significant relationship to posttest depression only in the major depressed group [R 2 change = .142, t(24) = 2.5, p < .05]. The


proportions of explained variance in the dysthymic, comorbid and agoraphobic group were .005, .065, and .055, respectively.

Regarding 1-year follow-up BDI scores, pretest BDI scores predicted a significant amount [R 2 change = .130, t(93) = 3.5, p < .001]. Pretest bad events globality ratings added a significant amount [R 2 change = .046, t(92) = 2.1, p < .05]. Within-group analyses revealed no significant relationships between globality and follow-up depression. The proportions of explained variance in the dysthymic, the major depressed, the comorbid, and the agoraphobic group were .003, .040, .015, and .028, respectively.

Posttest BDI scores predicted a significant amount of the variance in 1-year follow-up BDI scores [R 2 change = .191, t(92) = 4.4, p < .0001]. No posttest attributional variable added a significant amount in explained variance.

Stepwise regression with CPRS-D scores showed that pretest CPRS-D scores explained a significant amount of the CPRS-D posttest scores [R 2 change = .071, t(93) = 2.1,p < .05]. And again, pretest bad events globality ratings were the only attributional factor that sequentially added a significant amount in explained variance [R 2 change = .127, t(92) = 3.6, p < .0011.

And, finally, stepwise regression with the CPRS-A scores showed that pretest CPRS-A scores predicted a significant amount of the posttest CPRS-A scores [R 2 change = .528, t(93) = 9.6, p < .0001]. Thereafter, bad events globality scores predicted a close to significant amount [R 2 change = .020, t(92) = 1.9, p < .07].

DISCUSSION

The present results did not support the hypotheses that dysthymic and comorbid (major depressed and agoraphobic) patients would attribute bad events to more stable and global causes, and assign more importance to these events, than nonagoraphobic major depressed and nondepressed agoraphobic patients. Actually, the dysthymic patients did not score higher than the patients in the other groups on any of these attributional dimensions. As predicted, comorbid patients scored higher than the agoraphobic patients on the bad events stability and globality dimensions and higher than both the agoraphobic and the major depressed patients on the importance dimension. Various control analyses suggested that the attributional group differences were not mediated by differences in age, social class, presence of personality disorder, and treatment received. However, for every attributional difference obtained, there was a corresponding difference in level of depressive symptoms. When controlling statistically for


depressive symptom level, all group differences on the bad events attribution measures disappeared.

Consistent with Abramson and co-workers' (1989) view that cognitive biases will be more pronounced when the subject is in a depressed state, the scores on the bad events stability, globality, and importance measures decreased when depressive symptom level abated. The relationship between symptoms and attributions was further supported by the findings of significant partial correlations between most bad events attribution scores and concurrent depressive symptom scores, controlling for anxious symptoma- tology. Similar patterns were obtained across diagnostic groups at pretest. At posttest, attributions and depression seemed to be more closely related in the major depressed and the comorbid group than in the dysthymic and agoraphobic group. On the other hand, attributions were minimally related to anxiety symptoms.

However, influence does not appear to be directed only from symptoms to attributions. Supporting previous findings, attributional dimensions were found to predict later depression. Although the variance due to initial depression level was controlled for, bad events globality ratings at pretest accounted for posttest and follow-up depression scores. Contrary to our expectations, this predictive relationship seemed to be nonexistent or mar- ginal in the dysthymic and the comorbid group. In the major depressed group, the relation between bad events globality and later depression seemed to be more clear. The hypothesis that also bad events stability and importance scores should contribute independently to later depression was not supported. That the globality dimension turned out to be the stronger predictor is consistent with Peterson, Villanova, and Raps' (1985) review of studies of the relationship between attributions and concurrent depression. For the posttest depression level, significant results were obtained with both self-report and clinician-rated depression scales, indicating that the relationships between attribution and later depression were not artifacts resulting from a common response mode. Furthermore, attribution scores exhibited moderate stability in terms of relative level between individuals across time, whereas depression scores showed low stability. Together with the results of the predictive analyses, these results are consistent with the view that attributional variables may represent a fairly stable vulnerability factor present in many of the patients of this study. However, the applied DSM-III-R system of classifying patients was apparently not suited to iden- tify the vulnerable individuals.

Interpretation of the present results was limited by the failure to as- certain the presence of dysthymia in major depressed patients. Many of these patients might have been dysthymic as well, and this may have at- tenuated potential attributional differences between the present dysthymic


and m a j o r dep re s sed groups . It is also wor th not ing tha t to examine the

re la t ion be tween causal a t t r ibu t ions and la te r depress ion is not a d i rec t test o f the hope lessness theory (Brewin, 1985). Such a test would include the assessment of life events to see whe the r the occu r rence o f negat ive life events m o d e r a t e d the abil i ty of a t t r ibu t ions to predic t depress ion .

A C K N O W L E D G M E N T S

W e a re i ndeb ted to Prof. Svenn Torge r sen for supervis ing the pro jec t and to Thor l e i f L u n d for gu idance regard ing stat ist ical mat ters .

R E F E R E N C E S

Aasberg, M., Montgomery, S. A., Perris, C., Schalling, D., & Sedvall, G. (1978). A comprehensive psychopathological rating scale. Acta Psychiatrica Scandinavica Supplements, 271, 5-27.

Abramson, L. Y., Alloy, L. B., & Metalsky, G. I. (1989). Hopelessness depression: A theory-based subtype of depression. Psychological Review, 96, 358-372.

Alloy, L. B., Kelly, K. A., Mineka, S., & Clements, C. M. (1990). Comorbidity in anxiety and depressive disorders. A helplessness/hopelessness perspective. In J. D. Maser & C. R. Cloninger (eds.), Comorbidity of mood and anxiety disorders. Washington, DC: American Psychiatric Press.

American Psychiatric Association (1987). Diagnostic and statistical manual of mental disorders (3rd ed., rev.). Washington, DC: APA.

Beck, A. T., Steer, R. A., & Garbin, M. G. (1988). Psychometric properties of the Beck Depression Inventory: Twenty-five years of evaluation. Clhffcal Psychology Review, 8, 77-100.

Brewin, C. R. (1985). Depression and causal attribution: What is their relation? Psychological Bulletin, 98, 297-309.

Eaves, G., & Rush, A. J. (1984). Cognitive patterns in symptomatic and remitted unipolar major depression. Journal of Abnormal Psychology, 93, 31-40.

Hamilton, E. W., & Abramson, L. Y. (1983). Cognitive patterns and major depressive disorder: A longitudinal study in a hospital setting. Journal of Abnormal Psychology, 92, 173-184.

Heimberg, R. G., Klosko, J. S., Dodge, C. S., Shadick, R., Becker, R. E., & Barlow, D. H. (1989). Anxiety disorders, depression, and attributional style: A further test of the specificity of depressive attributions. Cognitive Therapy and Research, 13, 21-36.

Hoffart, A., & Martinsen, E. W. (1990a). Agoraphobia, depression, mental health locus of control, and attributional styles. Cognitive Therapy and Research, 14, 343-351.

Hoffart, A., & Martinsen, E. W. (1990b). Exposure based integrated vs. pure psychodynamic treatment of agoraphobic inpatients. Psychotherapy, 27, 210-218.

Martinsen, E. W., Friis, S., & Hoffart, A. (1989). A factor analytical study of the Comprehensive Psychopathological Rating Scale. Acta Psychiatrica Scandinavica, 80, 492-498.

Peterson, C., Semmel, A., von Baeyer, C., Abramson, L. Y., Metalsky, G. I., & Seligman, M. E. P. (1982). The Attributional Style Questionnaire. Cognitive Therapy and Research, 6, 287-299.

Peterson, C., Villanova, P., & Raps, C. S. (1985). Depression and attributions: Factors responsible for inconsistent results in the literature. Journal of Abnotznal Psychology, 94, 165-168.


Spitzer, R. L., & Williams, J. B. W. (1984). Structured Clinical Interview for DSM-III (SCID 10/1/84). Biometric Research Department, New York State Psychiatric Institute, 722 West 168th Street, New York 10032.

causal attributions in clinical subtypes of depression: a longitudinal study of inpatients

Documents