8/3/2019 36656BDDd01

1/6

BackgroundBackground Self-efficacy, aSelf-efficacy, a

characteristic thatis protective againstcharacteristic thatis protective against

depressive symptoms, maybedepressive symptoms, maybe

undermined by stressfullife events.undermined by stressfullife events.

AimsAims To estimate the effects of stressfulTo estimate the effects of stressful

life events on self-efficacy, andto examinelife events on self-efficacy, andto examineself-efficacyas a mediator ofthe effectofself-efficacy as a mediator ofthe effectof

stressfullife events on symptoms ofstressfullife events on symptoms of

depression.depression.

MethodMethod Using a sample of 2858Using a sample of 2858

respondents fromthe longitudinalrespondents fromthe longitudinal

Americans'Changing Lives study, pathAmericans'Changing Lives study, path

analyses wereused to evaluateanalyses were used to evaluate

interrelationships between self-eff icacy,interrelationships between self-eff icacy,

life events and symptoms of depressionlife events and symptoms of depression

controlling for a varietyof potentiallycontrolling for a variety of potentiallyconfounding variables. Separate modelsconfounding variables. Separatemodels

were estimated for those with andwere estimated for those with and

without priordepression.without priordepression.

ResultsResults For those with priorFor those with prior

depression, dependent life events had adepression, dependent life events had a

significant, negative impact on self-significant, negative impact on self-

efficacy.For those without priorefficacy.For those without prior

depression, life events had no effect ondepression, life events had no effecton

self-efficacy.self-efficacy.

ConclusionsConclusions For those with priorFor those with priordepression, self-efficacy mediatesdepression, self-efficacy mediates

approximately 40 % ofthe effect ofapproximately 40 % ofthe effect of

dependent stressfullife events ondependent stressfullife events on

symptoms of depression.symptoms of depression.

Declaration of interestDeclaration of interest SponsoredSponsored

by the Donaghue Women's Healthby the Donaghue Women's Health

Investigator Program at Yale University.Investigator Program at Yale University.

Since the introduction of the concept ofSince the introduction of the concept of

self-efficacy by Bandura (1977), studiesself-efficacy by Bandura (1977), studies

examining the relationship between self-examining the relationship between self-

efficacy and depression have found thatefficacy and depression have found that

those who score higher on measures ofthose who score higher on measures of

self-efficacy show substantially fewer symp-self-efficacy show substantially fewer symp-

toms of depression (Cutrona & Troutman,toms of depression (Cutrona & Troutman,

1986; McFarlane1986; McFarlane et alet al, 1995). However, as, 1995). However, aselaborated by Bandura (1997), self-efficacyelaborated by Bandura (1997), self-efficacy

is not a static characteristic. In theory, itis not a static characteristic. In theory, it

can be altered by behaviour, by internalcan be altered by behaviour, by internal

personal factors in the form of cognitive,personal factors in the form of cognitive,

affective and biological events, and by theaffective and biological events, and by the

external environment. This report providesexternal environment. This report provides

a more complete evaluation of the relation-a more complete evaluation of the relation-

ship between self-efficacy and depressionship between self-efficacy and depression

by examining the interrelationships be-by examining the interrelationships be-

tween self-efficacy, stressful life events andtween self-efficacy, stressful life events and

symptoms of depression in a longitudinalsymptoms of depression in a longitudinal

study using a large community sample. Instudy using a large community sample. In

particular, we tested the following hypo-particular, we tested the following hypo-theses: (a) symptoms of depression under-theses: (a) symptoms of depression under-

mine self-efficacy; (b) stressful life eventsmine self-efficacy; (b) stressful life events

undermine self-efficacy; and (c) self-efficacyundermine self-efficacy; and (c) self-efficacy

mediates the effect of stressful life events onmediates the effect of stressful life events on

symptoms of depression.symptoms of depression.

ContextContext

One of the aims of the present study was toOne of the aims of the present study was to

test the general validity of the notion thattest the general validity of the notion that

higher levels of self-efficacy result in fewerhigher levels of self-efficacy result in fewer

symptoms of depression. Prior studies lend-symptoms of depression. Prior studies lend-

ing support to this hypothesis haveing support to this hypothesis have usedusedcontext-specific measures of self-context-specific measures of self-efficacyefficacy

on narrowly defined populations perceivedon narrowly defined populations perceived

to be at risk for depression. Cutrona &to be at risk for depression. Cutrona &

Troutman (1986) examinedTroutman (1986) examined the relation-the relation-

ship between parenting self-ship between parenting self-efficacy andefficacy and

post-partum depression (post-partum depression (nn55). McFarlane55). McFarlane

et alet al (1995) studied the influence of social(1995) studied the influence of social

self-efficacy on depression in a study ofself-efficacy on depression in a study of

high-school students (high-school students (nn682). In contrast682). In contrast

to these studies, which explored the rela-to these studies, which explored the rela-

tionships between context-specific mea-tionships between context-specific mea-

sures of self-efficacy and depression insures of self-efficacy and depression in

narrowly defined study groups, we examin-narrowly defined study groups, we examin-ed the relationship between a globaled the relationship between a global

measure of personal efficacy and depressivemeasure of personal efficacy and depressive

symptoms, using a large sample of respon-symptoms, using a large sample of respon-

dents who participated in the Americans'dents who participated in the Americans'

Changing Lives (ACL) study.Changing Lives (ACL) study.

Over the past three decades, there hasOver the past three decades, there has

been compelling evidence for an associationbeen compelling evidence for an association

between stressful life events and depressionbetween stressful life events and depression(Brown & Harris, 1978; Surtees(Brown & Harris, 1978; Surtees et al et al ,,

1986; Kendler1986; Kendler et al et al , 1995). However,, 1995). However,

although the majority of those who becomealthough the majority of those who become

depressed have recently suffered a stressfuldepressed have recently suffered a stressful

life event, studies indicates that at leastlife event, studies indicates that at least

part of the association between stressfulpart of the association between stressful

life events and depression is non-causallife events and depression is non-causal

(Paykel, 1978; Kendler(Paykel, 1978; Kendler et alet al, 1999). In par-, 1999). In par-

ticular, some studies have suggested thatticular, some studies have suggested that

depression makes a person vulnerable todepression makes a person vulnerable to

the subsequent experience of stressful lifethe subsequent experience of stressful life

events (Hammen, 1991). In the presentevents (Hammen, 1991). In the present

study, we considered not only the possibil-study, we considered not only the possibil-ity that depressive symptoms predict certainity that depressive symptoms predict certain

types of stressful life events, but also thetypes of stressful life events, but also the

possibility that part of the effect of stressfulpossibility that part of the effect of stressful

life events on depressive symptoms islife events on depressive symptoms is

mediated through the impact of stressfulmediated through the impact of stressful

life events on self-efficacy.life events on self-efficacy.

The present study had three specificThe present study had three specific

aims. First, we tested whether higher levelsaims. First, we tested whether higher levels

of global self-efficacy would result in fewerof global self-efficacy would result in fewer

symptoms of depression. Second, we ex-symptoms of depression. Second, we ex-

amined the effects of symptoms of depres-amined the effects of symptoms of depres-

sion and stressful life events on self-sion and stressful life events on self-

efficacy. Third, we estimated the degree toefficacy. Third, we estimated the degree towhich the effect of stressful life events onwhich the effect of stressful life events on

symptoms of depression is mediatedsymptoms of depression is mediated

through self-efficacy.through self-efficacy.

METHODMETHOD

SampleSample

The ACL study was conducted by theThe ACL study was conducted by the

Survey Research Center of the UniversitySurvey Research Center of the University

of Michigan and we obtained the databaseof Michigan and we obtained the database

through the Inter-University Consortiumthrough the Inter-University Consortium

for Political and Social Research (House,for Political and Social Research (House,1994). The ACL is a multi-stage stratified1994). The ACL is a multi-stage stratified

area probability sample of people over thearea probability sample of people over the

age of 25 living throughout the continentalage of 25 living throughout the continental

United States. African Americans and thoseUnited States. African Americans and those

60 years of age and older were over-60 years of age and older were over-

sampled. Designed as a longitudinal studysampled. Designed as a longitudinal study

of productivity and successful aging in theof productivity and successful aging in the

middle and later years of life, the ACLmiddle and later years of life, the ACL

database includes measures relevant to thedatabase includes measures relevant to the

study of psychosocial influences onstudy of psychosocial influences on

depression.depression.

The ACL survey was conducted inThe ACL survey was conducted in

waves, with a baseline survey in 1986waves, with a baseline survey in 1986(Wave I) and a follow-up survey in 1989(Wave I) and a follow-up survey in 1989

3 7 33 7 3

B R I T I S H J O U R N A L O F P S Y C H I AT R YB R I T I S H J O U R N A L O F P S Y C H I AT R Y ( 2 0 0 0 ) , 1 7 6 , 3 7 3 ^ 3 7 8( 2 0 0 0 ) , 1 7 6 , 3 7 3 ^ 3 7 8

Self-efficacy as a mediator between stressful lifeSelf-efficacy as a mediator between stressful life

events and depressive symptomsevents and depressive symptoms

Differences based on history of prior depressionDifferences based on history of prior depression

PAUL K. MACIEJEWSKI, HOLLY G. PRIGERSON and CAROLYN M. MAZUREPAUL K. MACIEJEWSKI, HOLLY G. PRIGERSON and CAROLYN M. MAZURE

8/3/2019 36656BDDd01

2/6

M A C I E J E W S K I E T A LM A C I E J E W S K I E T A L

(Wave II). At baseline, a sample of 3617 re-(Wave II). At baseline, a sample of 3617 re-

spondents was interviewed in their homes.spondents was interviewed in their homes.

At follow-up, 2867 of these respondentsAt follow-up, 2867 of these respondents

who were interviewed at baseline were re-who were interviewed at baseline were re-

interviewed; they represented 83% of theinterviewed; they represented 83% of the

respondents at baseline who were still liv-respondents at baseline who were still liv-

ing at the time of the follow-up interview.ing at the time of the follow-up interview.Non-response did not vary significantly byNon-response did not vary significantly by

age, race or other known characteristics ofage, race or other known characteristics of

the respondents. Further information aboutthe respondents. Further information about

the ACL study is provided elsewherethe ACL study is provided elsewhere

(House(House et alet al, 1990). We focused on those, 1990). We focused on those

respondents in the ACL study interviewedrespondents in the ACL study interviewed

both at baseline and at follow-up for whomboth at baseline and at follow-up for whom

data on variables of interest were availabledata on variables of interest were available

((nn2858, or 99.7% of the 2867 respon-2858, or 99.7% of the 2867 respon-

dents interviewed at follow-up).dents interviewed at follow-up).

Because we hypothesised that the inter-Because we hypothesised that the inter-

relationships between self-efficacy, liferelationships between self-efficacy, life

events and symptoms of depression wouldevents and symptoms of depression woulddepend on whether there was a prior his-depend on whether there was a prior his-

tory of depression, the sample was dividedtory of depression, the sample was divided

into two groups: those not reportinginto two groups: those not reporting

((nn1610) and those reporting (1610) and those reporting (nn1248)1248)

at least one period when they felt sad, `blue'at least one period when they felt sad, `blue'

or depressed most of the time, or when theyor depressed most of the time, or when they

lost all interest and pleasure in things aboutlost all interest and pleasure in things about

which they usually care or enjoy. Thiswhich they usually care or enjoy. This

period was required to have lasted at leastperiod was required to have lasted at least

one week and have occurred prior to base-one week and have occurred prior to base-

line interview. Dividing up the sample inline interview. Dividing up the sample in

this way served to separate those withoutthis way served to separate those without

from those with a prior history of somefrom those with a prior history of someform of acute depression, but not necessa-form of acute depression, but not necessa-

rily major depression.rily major depression.

Measures for modelling depressiveMeasures for modelling depressive

symptomssymptoms

For each of the measures described below,For each of the measures described below,

`baseline' refers to data acquired at the Wave`baseline' refers to data acquired at the Wave

I (1986) interview; `follow-up' refers to dataI (1986) interview; `follow-up' refers to data

acquired at the Wave II (1989) interview.acquired at the Wave II (1989) interview.

Depressive symptomsDepressive symptoms

The severity of symptoms of depression wasThe severity of symptoms of depression was

assessed at baseline and follow-up, using aassessed at baseline and follow-up, using a

standardised measure of an 11-item shortstandardised measure of an 11-item short

form of the Center for Epidemiological Stu-form of the Center for Epidemiological Stu-

dies Depression Scale developed by Kohoutdies Depression Scale developed by Kohout

e t al e t al (1993). Kohout and his colleagues(1993). Kohout and his colleagues

found this 11-item version to be reliablefound this 11-item version to be reliable

(Cronbach's(Cronbach's aa0.81) and closely associated0.81) and closely associated

with the 20-item scale (with the 20-item scale (rr0.95).0.95).

Self-efficacySelf-efficacy

Personal beliefs about the ability to controlPersonal beliefs about the ability to control

one's environment and life circum-one's environment and life circum-stancesstances generally that is, one's globalgenerally that is, one's global

self-self-efficacy were assessed at baselineefficacy were assessed at baseline

and follow-up using a six-item standardisedand follow-up using a six-item standardised

index (Cronbach'sindex (Cronbach's aa0.67) representing a0.67) representing a

combination of Rosenberg's (1965) self-combination of Rosenberg's (1965) self-

esteem scale and Pearlin & Schooler'sesteem scale and Pearlin & Schooler's

(1978) mastery scale. This measure of per-(1978) mastery scale. This measure of per-

sonal efficacy consisted of items that aresonal efficacy consisted of items that aresimilar to those introduced by Sherer andsimilar to those introduced by Sherer and

colleagues and consistent with the conceptcolleagues and consistent with the concept

of self-efficacy as presented by Banduraof self-efficacy as presented by Bandura

(Bandura, 1977; Sherer(Bandura, 1977; Sherer et alet al, 1982)., 1982).

Stressful life eventsStressful life events

Measures of stressful life events were basedMeasures of stressful life events were based

on events occurring within a period of 12on events occurring within a period of 12

months prior to the follow-up interview.months prior to the follow-up interview.

Interviewers documented events using aInterviewers documented events using a

simple inventory, comprising: the death ofsimple inventory, comprising: the death of

a child, death of a spouse, death of a parent,a child, death of a spouse, death of a parent,

death of a close friend or relative, divorce,death of a close friend or relative, divorce,move to a new residence, loss of job, a ser-move to a new residence, loss of job, a ser-

ious financial problem, physical attack, andious financial problem, physical attack, and

life-threatening illness or injury. We fo-life-threatening illness or injury. We fo-

cused on these types of events because theycused on these types of events because they

have been found to be predictive of the on-have been found to be predictive of the on-

set of depression (Brown & Harris, 1978;set of depression (Brown & Harris, 1978;

KendlerKendler et alet al, 1995), and because they re-, 1995), and because they re-

present events considered severe in naturepresent events considered severe in nature

as assessed by patients and communityas assessed by patients and community

respondents alike (Grantrespondents alike (Grant et alet al, 1981)., 1981).

Because we hypothesised that self-Because we hypothesised that self-

efficacy would tend to be undermined byefficacy would tend to be undermined by

stressful life events over which the indivi-stressful life events over which the indivi-dual might have reasonably had some con-dual might have reasonably had some con-

trol, we divided the life events that wetrol, we divided the life events that we

considered into two categories; eventsconsidered into two categories; events

judged to be almost certainly independentjudged to be almost certainly independent

of the individual's behaviour, and eventsof the individual's behaviour, and events

judged to be at least partly dependent onjudged to be at least partly dependent on

it. This division of life events into theit. This division of life events into the

categories `independent' and `dependent' iscategories `independent' and `dependent' is

similar to that of Kendlersimilar to that of Kendler et alet al (1999). In(1999). In

the present study, the independent eventsthe present study, the independent events

documented included death of a childdocumented included death of a child

((nn19), death of a spouse (19), death of a spouse (nn30), death30), death

of a parent (of a parent (nn96), and death of a close96), and death of a closefriend or relative (friend or relative (nn660). Dependent660). Dependent

events included divorce (events included divorce (nn37), move to a37), move to a

new residence (new residence (nn314), loss of job314), loss of job

((nn87), serious financial problem87), serious financial problem

((nn216), physical attack (216), physical attack (nn13), and life-13), and life-

threatening illness or injury (threatening illness or injury (nn101). Sum-101). Sum-

mary measures were used to tally themary measures were used to tally the

number of stressful life events within eachnumber of stressful life events within each

of these two categories for each respondent.of these two categories for each respondent.

Control variablesControl variables

A variety of factors have been associatedA variety of factors have been associated

with depression, including social andwith depression, including social anddemographic factors (Kesslerdemographic factors (Kessler et alet al, 1994),, 1994),

chronic financial hardship (Brown & Mor-chronic financial hardship (Brown & Mor-

an, 1997), functional impairment (Zeissan, 1997), functional impairment (Zeiss etet

alal, 1996) and chronic health conditions, 1996) and chronic health conditions

(Black(Black et alet al, 1998). In order to ensure that, 1998). In order to ensure that

results in the present study would not beresults in the present study would not be

confounded with these factors, measuresconfounded with these factors, measures

for each were used as control variables infor each were used as control variables inour models for predicting stressful lifeour models for predicting stressful life

events, self-efficacy and symptoms of de-events, self-efficacy and symptoms of de-

pression. More specifically, respondents'pression. More specifically, respondents'

age, gender, race (Caucasian/non-Caucasian),age, gender, race (Caucasian/non-Caucasian),

socio-economic status, chronic financialsocio-economic status, chronic financial

stress, functional health status and numberstress, functional health status and number

of chronic health conditions (each assessedof chronic health conditions (each assessed

at baseline) were employed as control vari-at baseline) were employed as control vari-

ables. Socio-economic status was assessedables. Socio-economic status was assessed

using education and income level to classifyusing education and income level to classify

respondents into two categories represent-respondents into two categories represent-

ing lower and upper socio-economic status.ing lower and upper socio-economic status.

Chronic financial stress was determined byChronic financial stress was determined bymeans of a standardised index based on themeans of a standardised index based on the

work of Pearlin & Schooler (1978). Thiswork of Pearlin & Schooler (1978). This

index assessed the respondent's degree ofindex assessed the respondent's degree of

satisfaction with his/her present financialsatisfaction with his/her present financial

situation, degree of difficulty payingsituation, degree of difficulty paying

monthly bills, and ability to meet monthlymonthly bills, and ability to meet monthly

financial obligations. Functional healthfinancial obligations. Functional health

status was assessed by the ability to dostatus was assessed by the ability to do

heavy housework without difficulty, andheavy housework without difficulty, and

respondents were sorted into two categor-respondents were sorted into two categor-

ies, representing poor and good functionalies, representing poor and good functional

health. The number of chronic health con-health. The number of chronic health con-

ditions was assessed as the number of con-ditions was assessed as the number of con-ditions afflicting the respondent, andditions afflicting the respondent, and

included arthritis, lung disease, hyperten-included arthritis, lung disease, hyperten-

sion, heart attack, diabetes, cancer, footsion, heart attack, diabetes, cancer, foot

problems, stroke, broken bones and urinaryproblems, stroke, broken bones and urinary

incontinence.incontinence.

Analyses predicting symptomsAnalyses predicting symptoms

of depressionof depression

We calculatedWe calculated tt-statistics to test for differ--statistics to test for differ-

ences between those without and thoseences between those without and those

with prior depression (i.e. depressed for awith prior depression (i.e. depressed for a

period of at least one week at some timeperiod of at least one week at some timeprior to the baseline interview) in terms ofprior to the baseline interview) in terms of

symptoms of depression, self-efficacy andsymptoms of depression, self-efficacy and

number of life events. We calculatednumber of life events. We calculated

tt-statistics and-statistics and ww22--statistics to test forstatistics to test for

differences between these two groups withdifferences between these two groups with

respect to each of the control variables.respect to each of the control variables.

A path model was used to evaluate theA path model was used to evaluate the

direct and indirect effects of the baselinedirect and indirect effects of the baseline

level of symptoms of depression, self-level of symptoms of depression, self-

efficacy at baseline, number of independentefficacy at baseline, number of independent

life events, number of dependent life eventslife events, number of dependent life events

and self-efficacy at follow-up on symptomsand self-efficacy at follow-up on symptoms

of depression at follow-up. Path coefficientsof depression at follow-up. Path coefficientswere estimated separately for those withoutwere estimated separately for those without

3 743 74

8/3/2019 36656BDDd01

3/6

S E L F - E F F I C A C Y A N D P R I O R D E P R E S S I O NS E L F - E F F I C A C Y A N D P R I O R D E P R E S S I O N

and those with prior depression, for theand those with prior depression, for the

purpose of exploring aetiological aspectspurpose of exploring aetiological aspects

of the relationship between self-efficacy,of the relationship between self-efficacy,

life events and severity of symptoms of de-life events and severity of symptoms of de-

pression. Symptoms of depression at base-pression. Symptoms of depression at base-

line, self-efficacy at baseline, age, race,line, self-efficacy at baseline, age, race,

marital status, socio-economic status,marital status, socio-economic status,chronic financial stress, functional healthchronic financial stress, functional health

status and number of chronic health condi-status and number of chronic health condi-

tions were specified as exogenous variables.tions were specified as exogenous variables.

The number of independent life events,The number of independent life events,

number of dependent life events at self-number of dependent life events at self-

efficacy follow-up, and symptoms of depres-efficacy follow-up, and symptoms of depres-

sion at follow-up, were specified assion at follow-up, were specified as

endogenous variables. The sub-model forendogenous variables. The sub-model for

the number of independent life events con-the number of independent life events con-

sisted of paths from age and race. Thesisted of paths from age and race. The

sub-model for the number of dependent lifesub-model for the number of dependent life

events consisted of paths from self-efficacyevents consisted of paths from self-efficacy

at baseline, symptoms of depression atat baseline, symptoms of depression atbaseline, age, marital status, chronic finan-baseline, age, marital status, chronic finan-

cial stress, and number of chronic healthcial stress, and number of chronic health

conditions. The sub-model for self-efficacyconditions. The sub-model for self-efficacy

at follow-up consisted of paths from self-at follow-up consisted of paths from self-

efficacy at baseline, symptoms of depres-efficacy at baseline, symptoms of depres-

sion at baseline, number of independent lifesion at baseline, number of independent life

events, number of dependent life events,events, number of dependent life events,

age, marital status, socio-economic status,age, marital status, socio-economic status,

chronic financial stress, and number ofchronic financial stress, and number of

chronic health conditions. The model forchronic health conditions. The model for

symptoms of depression at follow-up con-symptoms of depression at follow-up con-

sisted of paths from symptoms of depres-sisted of paths from symptoms of depres-

sion at baseline, number of independentsion at baseline, number of independentlife events, number of dependent life events,life events, number of dependent life events,

self-efficacy at follow-up, age, race, maritalself-efficacy at follow-up, age, race, marital

status, socio-economic status, chronicstatus, socio-economic status, chronic

financial stress, functional health statusfinancial stress, functional health status

and number of chronic health conditions.and number of chronic health conditions.

The methods employed in the path ana-The methods employed in the path ana-

lyses were consistent with those describedlyses were consistent with those described

in established texts (Bollen, 1989; Loehlin,in established texts (Bollen, 1989; Loehlin,

1998). The overall fit of each path model1998). The overall fit of each path model

was assessed by means of its modelwas assessed by means of its model ww22, its, its

goodness of fit index adjusted for degreesgoodness of fit index adjusted for degrees

of freedom (AGFI), and its root meanof freedom (AGFI), and its root mean

square error of approximation (RMSEA).square error of approximation (RMSEA).We usedWe used tt-statistics to assess the signifi--statistics to assess the signifi-

cance of individual path coefficients withincance of individual path coefficients within

each model.each model.

RESULTSRESULTS

As shown in Table 1, those with priorAs shown in Table 1, those with prior

depression had significantly more severedepression had significantly more severe

symptoms of depression, lower levels ofsymptoms of depression, lower levels of

self-efficacy, and a greater number of de-self-efficacy, and a greater number of de-

pendent stressful life events, on average,pendent stressful life events, on average,

than those without prior depression. Thosethan those without prior depression. Thosewith prior depression were significantlywith prior depression were significantly

younger, more likely to be female, lessyounger, more likely to be female, less

likely to be married, and more likely tolikely to be married, and more likely to

have poor functional health than thosehave poor functional health than those

without prior depression. Those with priorwithout prior depression. Those with prior

depression also suffered significantly higherdepression also suffered significantly higher

levels of chronic financial stress and alevels of chronic financial stress and a

greater number of chronic health conditions.greater number of chronic health conditions.

Path coefficients for models predictingPath coefficients for models predicting

symptoms of depression at follow-up insymptoms of depression at follow-up in

samples without and with prior depressionsamples without and with prior depressionare presented in Table 2. For both of theseare presented in Table 2. For both of these

groups, the existence of symptoms of depres-groups, the existence of symptoms of depres-

sion at baseline and of self-efficacy at fol-sion at baseline and of self-efficacy at fol-

low-up had strong, significant, directlow-up had strong, significant, direct

effects on symptoms of depression at follow-effects on symptoms of depression at follow-

up. In particular, greater self-efficacy atup. In particular, greater self-efficacy at

follow-up was associated with less severefollow-up was associated with less severe

symptoms of depression. The number ofsymptoms of depression. The number of

independent stressful life events had a sig-independent stressful life events had a sig-

nificant, direct effect on symptoms ofnificant, direct effect on symptoms of

depression at follow-up for the group with-depression at follow-up for the group with-

out prior depression, but not for the groupout prior depression, but not for the group

with prior depression. The number of depen-with prior depression. The number of depen-dent stressful life events had a significant,dent stressful life events had a significant,

direct effect on symptoms of depression atdirect effect on symptoms of depression at

follow-up for both groups. Self-efficacy atfollow-up for both groups. Self-efficacy at

baseline had a significant, indirect effect onbaseline had a significant, indirect effect on

symptoms of depression at follow-up forsymptoms of depression at follow-up for

both the group without prior depressionboth the group without prior depression

((bbindirectindirect770.114, s.e.0.114, s.e.0.033,0.033, PP550.001)0.001)

and the group with prior depressionand the group with prior depression

((bbindirectindirect770.186, s.e.0.186, s.e.0.036,0.036, PP550.001).0.001).

The number of dependent stressful lifeThe number of dependent stressful life

events had a significant, negative, directevents had a significant, negative, directeffect on self-efficacy at follow-up for theeffect on self-efficacy at follow-up for the

group with prior depression, but not forgroup with prior depression, but not for

the group without prior depression. Self-the group without prior depression. Self-

efficacy and symptoms of depression atefficacy and symptoms of depression at

baseline both had strong, significant, directbaseline both had strong, significant, direct

effects on self-efficacy at follow-up for botheffects on self-efficacy at follow-up for both

groups. More severe symptoms of depres-groups. More severe symptoms of depres-

sion at baseline were associated with lowersion at baseline were associated with lower

self-efficacy at follow-up. Symptoms of de-self-efficacy at follow-up. Symptoms of de-

pression at baseline only had a significant,pression at baseline only had a significant,

direct effect on the number of dependentdirect effect on the number of dependent

stressful life events for the group with priorstressful life events for the group with prior

depression. Indices of overall model fitdepression. Indices of overall model fitwere excellent for the model for the groupwere excellent for the model for the group

3 7 53 7 5

Table 1Table 1 Comparison of study groups without and with prior depressionComparison of study groups without and with prior depression

Time ofassessmentTime ofassessment Prior depressionPrior depression11 Comparative testComparative test

MeasureMeasure No (No (nn1610)1610) Yes (Yes (nn1248)1248)

BaselineBaselineDichotomous measuresDichotomous measures nn (%)(%) nn (%)(%) ww22 d.f.d.f.

Gender (% female)Gender (% female) 976 (60.6%)976 (60.6%) 848 (67.9%)848 (67.9%) 16.35***16.35*** 11

Race (% caucasian)Race (% caucasian) 1047 (65.0%)1047 (65.0%) 855 (68.5%)855 (68.5%) 3.823.82 11

Marital status (% married)Marital status (% married) 956 (59.4%)956(59.4%) 664 (53.2%)664 (53.2%) 10.91***10.91*** 11

Socio-economic status (% upper)Socio- economic s tatus (% upper) 638 (39.6%)638 (39.6%) 517 (41.4%)517 (41.4%) 0.940.94 11

Functional health status (% good)Functional health status (% good) 1309 (81.3%)1309 (81.3%) 941 (75.4%)941 (75.4%) 14.63***14.63*** 11

Continuous measuresContinuous measures Mean (s.d.)Mean (s.d.) Mean (s.d.)Mean (s.d.) tt d.f.d.f.

AgeAge 54.5 (17.1)54.5 (17.1) 50.9 (16.9)50.9 (16.9) 5.70***5.70*** 28562856

Chronic financial stressChronic financial stress22 770.03 (1.01)0.03 (1.01) 0.19 (1.09)0.19 (1.09) 775.51***5.51*** 25652565

No. of chronic health conditionsNo. of chronic health conditions33

1.27 (1.31)1.27 (1.31) 1.45 (1.45)1.45 (1.45) 773.20**3.20** 25382538Self-efficacySelf-efficacy22 0.14 (0.95)0.14 (0.95) 770.20 (1.06)0.20(1.06) 8.69***8.69*** 25212521

Depressive symptomsDepressive symptoms33 770.16 (0.87)0.16 (0.87) 0.37 (1.15)0.37 (1.15) 7713.74***13.74*** 22582258

Follow-upFollow-up

Continuous measuresContinuous measures Mean (s.d.)Mean (s.d.) Mean (s.d.)Mean (s.d.) tt d.f.d.f.

No. of independent life eventsNo. of independent life events44 0.28 (0.47)0.28 (0.47) 0.29 (0.49)0.29 (0.49) 770.750.75 28562856

No. of dependent life eventsNo. of dependent life events44 0.22 (0.48)0.22 (0.48) 0.33 (0.62)0.33 (0.62) 775.53***5.53*** 22892289

Self-efficacySelf-efficacy22 0.08 (0.98)0.08 (0.98) 770.18 (1.10)0.18 (1.10) 6.50***6.50*** 25072507

Depressive symptomsDepressive symptoms33 770.13 (0.91)0.13 (0.91) 0.22 (1.12)0.22 (1.12) 778.86***8.86*** 23642364

****PP550.01; ***0.01; ***PP550.001.0.001.1.Depressed for a period of at least one week some time prior to the baseline interview.1. Depressed for a period of at least one week some time prior to the baseline interview.

2. Measure provided by Americans' Changing Lives (ACL).2. Measure provided by Americans' Changing Lives (ACL).3. ACL s tandardised 11-item Center for Epidemiological Studies Depression Scale (Kohout3. ACL standardised 11-item Center for Epidemiological Studies Depression Scale (Kohout etaletal, 1993)., 1993).4. Events occurring within 12 monthsprior to the follow-up interview.4. Events occurring within12 months prior to the follow-up interview.

8/3/2019 36656BDDd01

4/6

M A C I E J E W S K I E T A LM A C I E J E W S K I E T A L

without prior depression (without prior depression (ww22141419.12,19.12,

AGFIAGFI0.99, RMSEA0.99, RMSEA0.015) and for the0.015) and for the

model for the group with prior depressionmodel for the group with prior depression

((ww22141414.06, AGFI14.06, AGFI0.99, RMSEA0.99, RMSEA0.002).0.002).

DISCUSSIONDISCUSSION

Figure 1 was constructed from the results ofFigure 1 was constructed from the results of

our path analyses (presented in Table 2) toour path analyses (presented in Table 2) tofacilitate our discussion of the significantfacilitate our discussion of the significant

interrelationships between self-efficacy,interrelationships between self-efficacy,

stressful life events and depressive symp-stressful life events and depressive symp-

toms.toms.

The significance of self-efficacyThe significance of self-efficacy

in relation to depressionin relation to depression

In agreement with earlier studies reportingIn agreement with earlier studies reporting

significant effects of context-specific mea-significant effects of context-specific mea-

sures of self-efficacy on symptoms of depres-sures of self-efficacy on symptoms of depres-sion in narrowly defined populations, wesion in narrowly defined populations, we

found that a more global measure of personalfound that a more global measure of personal

efficacy was a strong, significant, negativeefficacy was a strong, significant, negative

predictor of symptoms of depression in apredictor of symptoms of depression in a

sample representative of the US population.sample representative of the US population.

For each of our study groups those withFor each of our study groups those with

and without prior depression higher levelsand without prior depression higher levels

of self-efficacy predicted less severe symp-of self-efficacy predicted less severe symp-toms of depression. Indeed, the impact oftoms of depression. Indeed, the impact of

self-efficacy on depression was so strong thatself-efficacy on depression was so strong that

the indirect effect of self-efficacy at baseline,the indirect effect of self-efficacy at baseline,

as mediated through self-efficacy at follow-as mediated through self-efficacy at follow-

up, had a significant impact on symptomsup, had a significant impact on symptoms

of depression assessed three years later, con-of depression assessed three years later, con-

trolling for the effects of a wide array oftrolling for the effects of a wide array of

potentially confounding factors.potentially confounding factors.

Our path models also suggest that thereOur path models also suggest that there

is a dynamic interplay between self-efficacyis a dynamic interplay between self-efficacy

and symptoms of depression, which oper-and symptoms of depression, which oper-

ates over time. In both of our study groups,ates over time. In both of our study groups,

greater self-efficacy at baseline significantlygreater self-efficacy at baseline significantlypredicts less serious symptoms of depres-predicts less serious symptoms of depres-

sion at follow-up, and more serious symp-sion at follow-up, and more serious symp-

toms of depression at baseline significantlytoms of depression at baseline significantly

predicts poorer self-efficacy at follow-up.predicts poorer self-efficacy at follow-up.

It appears that efforts to establish andIt appears that efforts to establish and

maintain a sense of control over one's lifemaintain a sense of control over one's life

and environment might serve to build aand environment might serve to build a

certain degree of resistance to subsequentcertain degree of resistance to subsequent

symptoms of depression, while periods ofsymptoms of depression, while periods of

depression might undermine these efforts.depression might undermine these efforts.

Self-efficacy as a mediatorSelf-efficacy as a mediator

of the effects of stressful life eventsof the effects of stressful life events

on depressionon depression

For individuals with prior depression, depen-For individuals with prior depression, depen-

dent stressful life events not only had a sig-dent stressful life events not only had a sig-

nificant, direct effect on their symptoms ofnificant, direct effect on their symptoms of

depression, but also had a significant, nega-depression, but also had a significant, nega-

tive effect on self-efficacy. Given that poorertive effect on self-efficacy. Given that poorer

self-efficacy strongly predicts more severeself-efficacy strongly predicts more severe

symptoms of depression, the total effect ofsymptoms of depression, the total effect of

dependent stressful life events on symptomsdependent stressful life events on symptoms

of depression (of depression (bbtotaltotal0.079) was the combi-0.079) was the combi-nation of the direct (nation of the direct (bbdirectdirect0.047) and indir-0.047) and indir-

ect (ect (bbindirectindirect0.032) effects. In other words,0.032) effects. In other words,

for those with prior depression, only 60%for those with prior depression, only 60%

of the total effect of dependent life eventsof the total effect of dependent life events

on symptoms of depression was direct, whileon symptoms of depression was direct, while

40% of the total effect was indirect,40% of the total effect was indirect,

mediated through the impact of dependentmediated through the impact of dependent

life events on self-efficacy. In contrast, forlife events on self-efficacy. In contrast, for

those without prior depression, dependentthose without prior depression, dependent

stressful life events did not have a significantstressful life events did not have a significant

effect on self-efficacy. Evidently, for theeffect on self-efficacy. Evidently, for the

group without prior depression, the effectgroup without prior depression, the effect

of dependent life events on symptoms ofof dependent life events on symptoms ofdepression was only direct.depression was only direct.

3 7 63 7 6

Table 2Table 2 Path coefficients for s tudy groups without and with prior depressionPath coefficients for studygroups without and with prior depression

ModelModel Prior depressionPrior depression11

Predictor variablePredictor variable NoNo22 ((nn1610)1610) YesYes33 ((nn1248)1248)

bb s.e.s.e. bb s.e.s.e.

Sub-modelfor no. of independentlife eventsSub-modelfor no.of independent life events

AgeAge 0.078**0.078** 0.0250.025 0.067*0.067* 0.0280.028

Race, caucasianRace, caucasian 770.0220.022 0.0250.025 770.069*0.0 69* 0.0280.028

Sub-model for no. ofdependent life eventsSub-modelfor no.of dependentlifeevents

Baseline self-efficacyBaseline self-efficacy 0.0400.040 0.0260.026 0.0220.022 0.0320.032

Baseline depressive symptomsBaseline depressive symptoms 0.0330.033 0.0270.027 0.10 1**0.10 1** 0.0340.034

AgeAge 770.227***0.227*** 0.0290.029 770.232***0.232*** 0.0330.033

Marital status, marriedMarital status, married 770.084***0.084*** 0.0240.024 770.062*0.062* 0.0280.028

Chronic financial stressChronic financial stress 0.192***0.192*** 0.0260.026 0.145***0.145*** 0.0300.030

No. of chronic health conditionsNo. of chronic health conditions 0.073*0.073* 0.0290.029 0.0090.009 0.0320.032

Sub-model for self-efficacyat follow-upSub-model for self-efficacyat follow-up

Baseline self-efficacyBaseline self-efficacy 0.371***0.371*** 0.0240.024 0.439***0.439*** 0.0280.028

Baseline depressive symptomsBaseline depressive symptoms 770.133***0.133*** 0.0240.024 770.106***0.106*** 0.0300.030

No. of independent life eventsNo. of independent life events 770.0090.009 0.0210.021 770.0170.0 17 0.0230.023

No. of dependent life eventsNo. of dependent life events 770.0140.014 0.0220.022 770.076**0.076** 0.0250.025

AgeAge 770.069*0.069* 0.0280.028 770.0220.022 0.0300.030

Marital status, marriedMarital status, married 770.0280.028 0.0230.023 770.051*0.051* 0.0250.025

Socio-economic status, upperSocio-economic status, upper 0.078**0.078** 0.0250.025 0.076**0.076** 0.0280.028

Chronic financial stressChronic financial stress 770.079**0.079** 0.0250.025 770.0150.015 0.0280.028

No. of chronic health conditionsNo. of chronic health conditions 770.081**0.0 81** 0.0260.026 770.096***0.096*** 0.0280.028

Model forsymptoms of depression at follow-upModel forsymptoms of depression at follow-up

Baseline depressive symptomsBaseline depressive symptoms 0.300***0.300*** 0.0230.023 0.288***0.288*** 0.0250.025

No. of independent life eventsNo. of independent life events 0.046*0.046* 0.0200.020 770.0060.006 0.0210.021

No. of dependent life eventsNo. of dependent life events 0.046*0.046* 0.0210.021 0.047*0.047* 0.0220.022

Follow-up self-efficacyFollow-up self-efficacy 770.308***0.308*** 0.0220.022 770.423***0.423*** 0.0230.023

AgeAge 770.0270.027 0.0270.027 770.065*0.065* 0.0280.028

Race, caucasianRace, caucasian 770.065**0.0 65** 0.0220.022 770.048*0.048* 0.0220.022

Marital status, marriedMarital status, married 770.066**0.0 66** 0.0220.022 0.0280.028 0.0230.023

Socio-economic status, upperSocio-economic status, upper 770.0410.041 0.0250.025 770.053*0.053* 0.0250.025

Chronic financial stressChronic financial stress 0.0440.044 0.0240.024 0.0130.013 0.0250.025

Functional health status, goodFunctional health status, good 770.0130.0 13 0.0230.023 770.062*0.062* 0.0250.025

No. of chronic health conditionsNo. of chronic health conditions 0.051*0.051* 0.0260.026 0.072**0.072** 0.02 70.027

**PP550.05; **0.05; **PP550.01; ***0.01; ***PP550.001.0.001.1.Depressed for a period of at least one week some time prior to the baseline interview.1.Depressed for a period of at least one week some time prior to the baseline interview.2. Indices of overallmodelfit:2. Indices of overallmodelfit: ww22141419.12; goodness of fit index adjustedfor degrees of freedom(AGFI)19.12; goodness of fit index adjustedfor degrees of freedom(AGFI)0.99; rootmean0.99; rootmeansquare error of approximation (RMSEA)square error of approximation (RMSEA)0.015.0.015.3. Indices of overall model fit:3. Indices of overallm odel fit: ww22141414.06; AGFI14.06; AGFI0.99; RMSEA0.99; RMSEA0.002.0.002.

8/3/2019 36656BDDd01

5/6

S E L F - E F F I C A C Y A N D P R I O R D E P R E S S I O NS E L F - E F F I C A C Y A N D P R I O R D E P R E S S I O N

Implications for the aetiologyImplications for the aetiology

of depressionof depression

The interrelationships between self-efficacy,The interrelationships between self-efficacy,symptoms of depression and stressful lifesymptoms of depression and stressful life

events for those with and without prior de-events for those with and without prior de-

pression differed in three notable ways (seepression differed in three notable ways (see

Fig. 1). First, consistent with the notionFig. 1). First, consistent with the notion

that stressful life events are more likely tothat stressful life events are more likely to

occur before first- or second-episode de-occur before first- or second-episode de-

pressions than prior to recurrent depres-pressions than prior to recurrent depres-

sions (Ezquiagasions (Ezquiaga et alet al, 1987), independent, 1987), independent

stressful life events had a significant effectstressful life events had a significant effect

on symptoms of depression only for theon symptoms of depression only for the

group without prior depression. For thosegroup without prior depression. For those

with prior depression, independent stressfulwith prior depression, independent stressful

life events had no effect on their symptoms.life events had no effect on their symptoms.Second, consistent with the notion thatSecond, consistent with the notion that

depression makes a person vulnerable todepression makes a person vulnerable to

experiencing subsequent stressful life eventsexperiencing subsequent stressful life events

(Hammen, 1991), more severe symptoms(Hammen, 1991), more severe symptoms

of depression at baseline significantlyof depression at baseline significantlypredicted greater numbers of dependentpredicted greater numbers of dependent

stressful life events for those with priorstressful life events for those with prior

depression. For those who had not suffereddepression. For those who had not suffered

prior depression, depressive symptoms atprior depression, depressive symptoms at

baseline had no effect on the occurrencebaseline had no effect on the occurrence

of dependent stressful life events. Third, asof dependent stressful life events. Third, as

already noted, dependent stressful lifealready noted, dependent stressful life

events had a significant, negative effect onevents had a significant, negative effect on

self-efficacy for those who had sufferedself-efficacy for those who had suffered

prior depression. For those without priorprior depression. For those without prior

depression, dependent stressful life eventsdepression, dependent stressful life events

had no effect on self-efficacy.had no effect on self-efficacy.

Taken together, the findings of the pre-Taken together, the findings of the pre-sent study suggest a spiralling cycle ofsent study suggest a spiralling cycle of

depression. The cycle begins with a stressfuldepression. The cycle begins with a stressful

life event, either independent of, or depen-life event, either independent of, or depen-

dent on, the individual's behaviour, trigger-dent on, the individual's behaviour, trigger-

ing depression in someone with low self-ing depression in someone with low self-

efficacy. This episode of depression makesefficacy. This episode of depression makes

the individual vulnerable to experiencingthe individual vulnerable to experiencing

subsequent dependent stressful life events,subsequent dependent stressful life events,which serve to undermine their self-efficacywhich serve to undermine their self-efficacy

further. This, in turn, makes them vulner-further. This, in turn, makes them vulner-

able to subsequent depression, which pre-able to subsequent depression, which pre-

disposes to additional dependent stressfuldisposes to additional dependent stressful

life events, which continue to underminelife events, which continue to undermine

their self-efficacy. The result is yet anothertheir self-efficacy. The result is yet another

episode of depression, andthecyclecontinues.episode of depression, andthecyclecontinues.

Future directionsFuture directions

The impact of dependent stressful lifeThe impact of dependent stressful life

events on self-efficacy and depressionevents on self-efficacy and depression

among those with prior depression may beamong those with prior depression may be

closely related to the effects of explanatoryclosely related to the effects of explanatory

style on depression (Peterson & Seligman,style on depression (Peterson & Seligman,

1984). If a connection can be demonstrated1984). If a connection can be demonstrated

between style of causal attribution and self-between style of causal attribution and self-

efficacy in response to dependent stressfulefficacy in response to dependent stressful

life events, then efforts to build an optimis-life events, then efforts to build an optimis-

tic explanatory style may prove to be antic explanatory style may prove to be an

effective means of maintaining higher levelseffective means of maintaining higher levels

of self-efficacy in response to these events.of self-efficacy in response to these events.

In this case, such efforts might diminishIn this case, such efforts might diminish

the psychological consequences of depen-the psychological consequences of depen-

dent life events and reduce the risk of sub-dent life events and reduce the risk of sub-

sequent depression. This may prove to besequent depression. This may prove to be

particularly important for women, whomparticularly important for women, whom

we found to be significantly more likely towe found to be significantly more likely to

have had prior depression, and who hadhave had prior depression, and who had

significantly lower levels of self-efficacy.significantly lower levels of self-efficacy.

REFERENCESREFERENCES

Bandura, A. (1977)Bandura, A. (1977) Self-efficacy: toward a unifiedSelf-efficacy: toward a unified

theory of behavioral change.theory of behavioral change. Psychological ReviewPsychological Review,, 8484,,

191^215.191^215.

__(1997)(1997) Self-Efficacy: The Exercise of ControlSelf-Efficacy: The Exercise of Control. New. New

York: Freeman.York: Freeman.

Black, S. A., Goodwin, J. S. & Markides, K. S. (1998)Black, S. A., Goodwin, J. S. & Markides, K. S. (1998)

The association between chronic diseases andThe a ssociation between chronic diseases anddepressive symptomatology in older Mexicandepressive symptomatology in older Mexican

Americans.Americans.Journal of Gerontology. Series A, Biological andJournal of Gerontology. Series A, Biological and

Medical SciencesMedical Sciences,, 5353, 188^194., 188^194.

Bollen, K. A. (1989)Bollen, K. A. (1989) Structural Equations with LatentStructural Equations with Latent

VariablesVariables. NewYork: Wiley.. New York: Wiley.

Brown, G.W. & Harris,T.O. (1978)Brown, G.W. & Harris,T.O. (1978) Social Origins ofSocial Origins of

DepressionDepression. NewYork: Free Press.. NewYork: Free Press.

__& Moran, P. M. (1997)& Moran, P. M. (1997) Single mothers, poverty andSingle mothers, poverty and

depression.depression. Psychological MedicinePsychological Medicine,, 2727, 21^33., 21^33.

Cutrona, C. E. & Troutman, B. R. (1986)Cutrona, C. E. & Troutman, B. R. (1986) SocialSocial

support, infant temperament, and parenting self-support, infant temperament, and parenting self-

efficacy: a mediational model of postpartum depression.efficacy: a mediational model of postpartum depression.

Child DevelopmentChild Development,, 5757, 1507^1518., 1507^1518.

Ezquiaga, E., Gutierrez, J. L. A. & Lopez, A. G.Ezquiaga, E., Gutierrez, J. L. A. & Lopez, A. G.

(1987)(1987) Psychosocial factors and episode number inPsychosocial factors and episode number in

depression.depression. Journal of Affective DisordersJournal of Affective Disorders,, 1212, 135^138., 135^138.

3 7 73 7 7

Fig. 1Fig. 1 Interrelationships between self-efficacy, life events and depressive symptoms for study groups (a)Interrelationships between self-efficacy, life events and depressive symptoms for study groups (a)

without and (b) with prior depression (depressed for a period of at least one week some time prior to thewithout and (b) withprior depression (depressed for a period of at least one week some time prior to the

baseline interview).baseline interview).

**PP550.05, **0.05, **PP550.01, ***0.01, ***PP550.001. Only statistically significant paths are shown.Complete results are0.001. Only statistically significant paths are shown.Complete results are

presented inTable 2.presented inTable 2.

(a) Study group without prior depression(a) Study group without prior depression

(b) Study group with prior depression(b) Study group with prior depression

8/3/2019 36656BDDd01

6/6

M A C I E J E W S K I E T A LM A C I E J E W S K I E T A L

Grant, I., Hervey, L., Sweetwood, M. S.,Grant, I., Hervey, L., Sweetwood, M. S., et alet al (1981)(1981)

Quality of life events in relationto psychiatric symptoms.Quality of life events in relationto psychiatric symptoms.

Archives of General PsychiatryArchives of General Psychiatry,, 3838, 335^339., 335^339.

Hammen,C . (1991)Hammen, C. (1991) Generation of stress in the courseGeneration of stress in the course

of unipolar depression.of unipolar depression.Journal of Abnormal PsychologyJournal of Abnormal Psychology,,

100100, 555^561., 555^561.

House, J. S. (1994)House, J. S. (1994) Americans'Changing Lives: Waves IAmericans'Changing Lives: Waves I

and II, 1986 and1989and II, 1986 and1989 (computer file). ICPSRversion. Ann(computer file). ICPSRversion. Ann

Arbor,MI: Inter-University Consortium for Political andArbor,MI: Inter-University Consortium for Political and

Social Research.Social Research.

__, Kessler, R. C., Herzog, A. R.,, Kessler, R. C., Herzog, A. R., et alet al (1990)(1990) Age,Age,

socioeconomic status, and health.socioeconomic status, and health. Milbank QuarterlyMilbank Quarterly,, 6868,,

383^411.383^411.

Kendler, K. S., Kessler, R. C.,Walters, E. E.,Kendler, K. S., Kessler, R. C.,Walters, E. E., et alet al

(1995)(1995) Stressful life events, genetic liability, and onset ofStressfullife events, genetic liability, and onset of

a major depressive episode in women.a major depressive episode in women. American JournalAmerican Journal

of Psychiatryof Psychiatry,, 152152, 833^842., 833^842.

__, Karkowski, L. M. & Prescott, C. A. (1999), Karkowski, L. M. & Prescott, C. A. (1999) CausalCausal

relationship between stressful life events and the onsetrelationship between stressful life events and the onset

of a major depression.of a major depression. American Journal of PsychiatryAmerican Journal of Psychiatry,,

156156, 8 37^841., 837 ^841.

Kessler, R. C., McGonagle, K. A., Zhao, S.,Kessler, R. C., McGonagle, K. A., Zhao, S., et alet al

(1994)(1994) Lifetime and 12-month prevalence of DSM^ III ^ RLifetime and 12-month prevalence of DSM^ III ^ R

psychiatric disorders in the United States.psychiatric disorders in the United States. Archives ofArchives of

General PsychiatryGeneral Psychiatry,, 5151, 8^19., 8^19.

Kohout, F. J., Berkman, L. F., Evans, D. A.,Kohout, F. J., Berkman,L. F., Evans, D. A., et alet al

(1993)(1993) Two shorter forms of the CES-D depressionTwo shorter forms of the CES-D depression

symptoms index.symptoms index.Journal of Aging & HealthJournal of Aging & Health,, 55, 179^193., 179^193.

Loehlin, J. C. (1998)Loehlin, J. C. (1998) Latent Variable Models: AnLatent Variable Models: An

Introduction to Factor, Path, and Structural AnalysisIntroduction to Factor, Path, and Structural Analysis (3rd(3rd

edn). Mahwah, NJ:Erlbaum.edn). Mahwah, NJ: Erlbaum.

McFarlane, A. H., Bellissimo, A. & Norman, G. R.McFarlane, A. H., Bellissimo, A. & Norman, G. R.

(1995)(1995) The role of family and peersin social self-efficacy:The role of family and peersin social self-efficacy:

links to depression in adolescence.links to depression in adolescence. American Journal ofAmerican Journal of

OrthopsychiatryOrthopsychiatry,, 6565, 402^410., 402^410.

Paykel, E. S. (1978)Paykel, E. S. (1978) Contribution of life events toContribution of life events tocausation of psychiatric illness.causation of psychiatric illness. Psychological MedicinePsychological Medicine,, 88,,

245^253.245^253.

Pearlin, L. I. & Schooler, C. (1978)Pearlin, L. I. & Schooler, C. (1978) The structure ofThe structure of

coping.coping.Journal of Health and Social BehaviorJournal of Health and Social Behavior,, 1919, 2^21., 2^21.

Peterson, C. & Seligman, M. E. (1984)Peterson, C. & Seligman, M. E. (1984) CausalCausal

explanations as a risk factor for depression.explanations as a risk factor for depression. PsychologicalPsychological

ReviewReview,, 9191, 347^374., 347^374.

Rosenberg, M. (1965)Rosenberg, M. (1965) Society and the Adolescent Self-Society and the Adolescent Self-

ImageImage. Princeton, NJ: Princeton University Press.. Princeton, NJ: Princeton University Press.

Sherer, M.,Maddux, J. E., Mecadante, B.,Sherer, M.,Maddux, J. E., Mecadante, B., et alet al (1982)(1982)

The self-efficacy scale: construction and validation.The self-efficacy scale: construction and validation.

Psychological ReportsPsychological Reports,, 5151, 663^671., 663^671.

Surtees, P. G., Miller, P. M.,Ingham, J. G.,Surtees, P. G., Miller, P. M.,Ingham, J. G., et alet al (1986)(1986)

Life events and the onset of affective disorder: aLife events and the onset of affective disorder: a

longitudinal general population study.longitudinal general population study.Journal of AffectiveJournal of Affective

DisordersDisorders,, 1010, 37^50., 37^50.

Zeiss, A. M., Lewinsohn,P.M., Rohde, P.,Zeiss, A. M.,Lewinsohn, P. M., Rohde, P., et alet al (1996)(1996)

Relationship of physical disease and functionalRelationship of physical disease and functional

impairment to depression in older pe ople.impairment to depression in older people. PsychologyPsychology

and Agingand Aging,, 1111, 5 72^581., 572^581.

3 7 83 7 8

CLINICAL IMPLICATIONSCLINICAL IMPLICATIONS

&& Those with low self-efficacy are at risk for severe symptoms of depression.Those with low self-efficacy are at risk for severe symptoms of depression.

&& Efforts to establish and maintain higher levels of self-efficacy may serve to build upEfforts to establish and maintain higher levels of self-efficacy may serve to build up

a long-term resistance to symptoms of depression in the future.a long-term resistance to symptoms of depression in the future.

&& For those with prior depression, efforts to enhance self-efficacy following stressfulFor thosewithprior depression, efforts to enhance self-efficacy following stressful

life events, perhaps through cognitive^behavioural psychotherapeutic techniques,life events, perhaps through cognitive^ behavioural psychotherapeutic techniques,

may reduce the severity of subsequent symptoms of depression.may reduce the severity of subsequent symptoms of depression.

LIMITATIONSLIMITATIONS

&& The design of the study was restricted to the number (two) and timing (threeThe design of the study was restricted to the number (two) and timing (three

years' separation) of the waves of interviews available in the Americans' Changingyears' separation) of the waves of interviews available in the Americans' Changing

Lives study.Lives study.

&& The assessment of life events was restricted to a simple inventory.The assessment of life events was restricted to a simple inventory.

&& The results are restricted to symptoms of depression.The results are restricted to symptoms of depression.

PAUL K. MACIEJEWSKI, PhD, Donaghue Women's Health Investigator Program, Yale University School ofPAUL K. MACIEJEWSKI, PhD, Donaghue Women's Health Investigator Program, Yale University School of

Medicine,New Haven,CT; HOLLY G. PRIGERSON, PhD,Donaghue Women's Health Investigator Program andMedicine,New Haven,CT; HOLLY G. PRIGERSON, PhD,Donaghue Women's Health Investigator Program and

Department of Psychiatry,Yale University School of Medicine, New Haven, CT, and Mental Illness,ResearchDepartment of Psychiatry,Yale University School of Medicine, New Haven, CT, and Mental Illness,Research

and Clinical Center,Veterans Affairs Medical Center,West Haven, CT; CAROLYN M. MAZURE,PhD, Donaghueand Clinical Center,Veterans Affairs Medical Center,West Haven, CT; CAROLYN M. MAZURE,PhD, Donaghue

Women's Health Investigator Program and Depar tment of Psychiatry,Yale University Scho ol of Medicine, NewWomen's Health Investigator Program and Depar tment of Psychiatry,Yale University Scho ol of Medicine, New

Haven,CT,USAHaven,CT,USA

Correspondence:Dr Paul K. Maciejewski,Donaghue Women's Health Investigator Program,YaleCorrespondence:Dr Paul K. Maciejewski,Donaghue Women's Health Investigator Program,Yale

University School of Medicine,PO Box 20 80 91, New Haven,CT06520,USAUniversity School of Medicine, PO Box 20 80 91,New Haven,CT0 6520,USA

(First received18 June 1999, f inal revision1 October 1999, accepted1 October 1999)(First received18 June 1999, f inal revision1 October 1999, accepted 1 October 1999)