Available online at www.sciencedirect.com

General Hospital Psychiatry 33 (2011) 224–231

Depressive vulnerabilities predict depression status and trajectories ofdepression over 1 year in persons with acute coronary syndrome☆

Frank Doyle, Ph.D.a,⁎, Hannah McGee, Ph.D.a, Mary Delaney, M.A.b,Nicola Motterlini, StatSci.D.c, Ronán Conroy, D.Sc.d

aDivision of Population Health Sciences (Psychology), Royal College of Surgeons in Ireland, 123 St Stephen's Green, Dublin 2, IrelandbHRB Centre for Health and Diet Research, Department of Food Business and Development, University College Cork, Cork, Ireland

cHRB Centre for Primary Care Research, Division of Population Health Sciences, Royal College of Surgeons in Ireland,123 St Stephen's Green, Dublin 2, Ireland

dDivision of Population Health Sciences (Epidemiology & Public Health Medicine), Royal College of Surgeons in Ireland, 123 St Stephen's Green,Dublin 2, Ireland

Received 12 January 2011; accepted 9 March 2011

Objective: Depression is prevalent in patients hospitalized with acute coronary syndrome (ACS). We determined whether theoreticalvulnerabilities for depression (interpersonal life events, reinforcing events, cognitive distortions, Type D personality) predicted depression, ordepression trajectories, post-hospitalization.Methods: We followed 375 ACS patients who completed depression scales during hospital admission and at least once during three follow-up intervals over 1 year (949 observations). Questionnaires assessing vulnerabilities were completed at baseline. Logistic regression forpanel/longitudinal data predicted depression status during follow-up. Latent class analysis determined depression trajectories. Multinomiallogistic regression modeled the relationship between vulnerabilities and trajectories.Results: Vulnerabilities predicted depression status over time in univariate and multivariate analysis, even when controlling for baselinedepression. Proportions in each depression trajectory category were as follows: persistent (15%), subthreshold (37%), never depressed(48%). Vulnerabilities independently predicted each of these trajectories, with effect sizes significantly highest for the persistentdepression group.Conclusions: Self-reported vulnerabilities — stressful life events, reduced reinforcing events, cognitive distortions, personality — measuredduring hospitalization can identify those at risk for depression post-ACS and especially those with persistent depressive episodes.Interventions should focus on these vulnerabilities.© 2011 Elsevier Inc. All rights reserved.

Keywords: Depression; Coronary heart disease; Psychological theory; Life events; Personality; Just world beliefs

Depression is prevalent in patients with coronary heartdisease, with the prevalence estimated at approximately20% in patients with myocardial infarction [1]. This issignificantly higher than that seen in general populationsamples [2]. The importance of depression is highlightednot only in its prevalence, and its impact on quality of life,but also on the ability of depression to predict cardiovas-cular prognosis [3–5].

☆ Funded by the Health Research Board.⁎ Corresponding author. Tel.: +353 1 4022718; fax: +353 1 4022764.E-mail address: fdoyle4@csi.ie (F. Doyle).

0163-8343/$ – see front matter © 2011 Elsevier Inc. All rights reserved.doi:10.1016/j.genhosppsych.2011.03.008

However, while a large literature concerns the predictionof prognosis in depressed cardiac patients, relatively littleresearch is concerned with what happens to depression afterthe acute hospitalization phase. Depression is a chronic,episodic condition, and, therefore, research on whathappens to depressive symptoms in the post-acute phasepotentially provides vital information for interventiondesign. While the prevalence of depression is comparative-ly steady over time, this masks the different trajectoriessymptoms of depression take [6–8]. Indeed, sophisticatedstudies have shown different patterns of resolving persistentdepression in patients with heart disease [7,8]. For example,Martens et al. [7] surveyed 287 patients post-hospitalizationfor myocardial infarction at 2 and 12 months. They

225F. Doyle et al. / General Hospital Psychiatry 33 (2011) 224–231

categorized four groups of patients in relation to depressivesymptom status: nondepressed, mildly depressed, moder-ately depressed and severely depressed. Similarly, Kapteinet al. [8] followed 475 patients with myocardial infarctionevery 3 months over 1 year, and their results showed fivedistinct trajectories regarding depression: no depressivesymptoms, mild depressive symptoms, moderate andincreasing depressive symptoms, significant but decreasingdepressive symptoms and significant and increasingdepressive symptoms. Thus, the evolution of depressionis complex, and in order to design optimal interventions,more knowledge on the predictors of depressive symptomsand such depressive trajectories is needed [9].

While some research has established predictors ofdepression in patients with coronary heart disease fromeasily available variables recorded as part of standard hospitalcare, the results are often contradictory [7,8,10–12]. Forexample, age, sex, medications and left ventricular functionhave been shown to predict depression in cardiac patients insome of these findings, but not in others. Furthermore, suchfindings are atheoretical and thus provide little clue as to howto intervene in such populations [9,13]. A paucity of evidenceexists assessing the relative importance of theoreticalvulnerabilities, and their associated interventions, regardingrisk of depression and trajectories of depression after acutecoronary syndrome (ACS) [14]. While a small number ofstudies have assessed theoretical vulnerabilities to depres-sion — for example, stressful life events, personality andcognitions have all been associated with depression in cardiacpatients [7,15,16] — such studies have not measured thesevulnerabilities simultaneously or have not assessed theirassociation with trajectories of depression post-ACS.

These vulnerabilities are especially important, given recentfindings which suggest that, in patients with ACS, suchvulnerabilities predict depression better than do demographicor disease variables [13,17]. However, both these studies werelimited, as they were cross-sectional and did not allow for thedirection of causality to be determined [13,17]. Also, it waspossible that recall bias in depressed patients contributed to ahigher self-reported level of such vulnerabilities — thus toinflated correlations between the variables.We therefore reporton longitudinal data from our cohort. We aimed to determine(a) whether depressive vulnerabilities predicted depressionover time, when controlling for baseline depression, and (b)whether these vulnerabilities better predicted different types ofdepression (e.g., persistent depression).

1. Methods

1.1. Study design and participants

The baselinemethods have been reported previously [3,13].This article presents data from ACS patients who completeddepression questionnaires at baseline (during acute hospitaladmission) and who responded to at least one of the postalfollow-up surveys at 3, 6 and 12 months (not all participants

completed all theoretical vulnerability scales). Briefly, afterethical approval was provided, patients were recruited from 12hospitals. Consecutive patients with confirmed ACS (myo-cardial infarction or unstable angina) who were literate inEnglish were recruited by coronary care staff to participate inthe survey during their hospital stay. Patients completed acomposite psychological questionnaire while in-hospital, andcoronary disease risk factor and treatment data was obtainedfrom medical charts. Major co-morbidities were also recordedas per the Charlson Comorbidity Index [18] and modified byomitting some of the risk factors which are separately assessedin cardiac patients (e.g., myocardial infarction, diabetes).Patients were then followed up by postal survey, containingmeasures of depression, at each of the following three phases.Non-respondents were posted a reminder after 2 weeks andthen telephoned with a further reminder if no response wasreceived after another 2 weeks.

1.2. Measures

1.2.1. Depression scales

1.2.1.1. Beck Depression Inventory–Fast Screen. TheBeck Depression Inventory–Fast Screen (BDI-FS) is aseven-item scale focusing on cognitive symptoms ofdepression [19] and has very good sensitivity/specificity(N0.90/N0.85) for detecting major depression when using athreshold score of N3 [20,21]. We omitted the suicidalityitem, but maintained the threshold of N3, for reasons outlinedpreviously [3,6]. Also, the predictive power of the BDI-FShas been shown to be unchanged when removing this item inpersons with hepatitis C [22].

1.2.1.2. Hospital Anxiety and Depression Scale–Depressionsubscale. The Hospital Anxiety and Depression Scale(HADS) is a 14-item measure that was developed to measureanxiety and depression in hospitalized patients and omitssomatic items so scores are not contaminated by symptomsof chronic conditions [23]. We used the seven-item HADSDepression subscale (HADS-D) only and adopted therecommended threshold of N7 [24]. The HADS-D focusesmainly on anhedonia.

Scoring above threshold on either scale was considered toindicate depression status at baseline and follow-up.

1.2.2. Depressive vulnerability measures

1.2.2.1. List of Threatening Experiences Question-naire. Stressful interpersonal and life events (e.g., seriousillness or assault, or a relationship break-up) were assessedusing the 12-item List of Threatening Experiences Ques-tionnaire (LTE-Q) [25,26]. This schedule relates to eventsthat have happened in the prior year. The authors showedthat the LTE-Q had high test–retest reliability and comparedwell with an interview technique (sensitivity/specificityranges for stressful life events were between 0.89 and 1.0/between 0.74 and 0.88, respectively), in psychiatric patients.

226 F. Doyle et al. / General Hospital Psychiatry 33 (2011) 224–231

1.2.2.2. Pleasant Events Schedule–Alzheimer's Disease(short version). Pleasant events were assessed using thePleasant Events Schedule–Alzheimer's Disease (shortversion) (PES-AD), a 20-item behavioral log. The scalewas originally developed for persons with Alzheimer'sdisease [27], but has also been used in ACS patients [17].Environmental engagement is measured by ratings of thefrequency of behaviors/events, and enjoyment of same, inthe past month. A cross-product produces a total schedulescore of positive reinforcement in the past month. Missingitems were coded as zero if at least half of the 20 items hadbeen answered [17].

1.2.2.3. Belief in a Just World–Self scale. As a period ofadjustment postevent is likely for all patients, and not onlyfor those who have distorted cognitions or dysfunctionalattitudes or distorted cognitions, we assessed just-worldbeliefs instead of other types of cognitive distortions [13].Belief in a Just World–Self scale (BJW) refers to the beliefthat good things happen to good people and bad thingshappen to bad people [28], and a ‘distorted’ BJW (i.e.,nonbelief in a just world) has been associated withdepression [13,28,29]. BJW for self was assessed by theeight-item BJW-S [29].

1.2.2.4. Comparing vulnerabiltiesFor comparability among measures, and in line with

previous research [13,17], the scores of the vulnerabilityscales above were recoded to indicate a higher risk fordepression (i.e., a lack of positive reinforcement, notbelieving in a just world, but higher numbers of stressfullife events). For effect size comparability, scale scores werestandardized, with effect sizes representing a 1-S.D. increase.

1.2.2.5. Type D scale–DS14. The distressed (Type D)personality— a combination of both negative affectivity andsocial inhibition — was assessed using the 14-item DS14scale [30]. Scoring above threshold (N10) on both of thesubscales indicates those of Type D disposition. The DS14has been used extensively in cardiac patients, and it hasdemonstrated good psychometric properties [30].

1.2.3. StatisticsDifferences between groups were assessed with χ2 test or

analysis of variance as appropriate. Missing data wasimputed for depression and vulnerability scales using Stata'sregression-based ‘impute’ command [3,13], but imputationwas inappropriate for the schedules (LTE-Q, PES-AD).Logistic regression with random effects estimates for panel/longitudinal data was adopted to allow prediction ofdepression status (person status) throughout the follow-upperiod, adjusting for baseline depression. Odds ratios (ORs)were used as a measure of effect size. Latent class analysis ofcombined HADS-D and BDI-FS score was conducted usingthe SAS PROC TRAJ command, as in previous research[31]. Adding age, sex, prior coronary heart disease and lowleft ventricular function as covariates had a negligible effect

on the depression trajectories, so the nonadjusted groupswere used in subsequent analysis. The lowest BayesianInformation Criterion value (−5449.68) led to one categorywith ∼3% of participants, so the next lowest was chosen(−5438.3). Both panel-modeling logistic regression andlatent class analysis are designed to account for missing dataduring follow-up. Multinomial logistic regression, reportingrelative risk ratios for effect sizes, was then used to model therelationship between vulnerabilities and different categoriesof depression during follow-up, using never depressed as thereference category. Post hoc Wald test statistics examinedwhether the effect sizes were significantly different for eachvulnerability when predicting depression categories.

2. Results

2.1. Response rate

During follow-up, 375 (87%) of 430 patients respondedto at least one of the follow-up surveys and 250 (58%) of 430patients responded to all of the follow-up surveys. Thisprovided, depending on the response to a particular scale atbaseline, up to 949 unique observations in the data. Non-respondents to the follow-up phases were less likely to havea partner (OR=0.6, 95% CI 0.4–0.9, P=.014) and less likelyto have private health insurance (OR=0.3, 95% CI 0.1–0.7,P=.005), but no other demographic differences were found.

2.2. Depression trajectories

The depression trajectories of the combined scales areshown in Fig. 1.

Numbers/proportions in each depression trajectory cate-gory were as follows: persistent— 57 (15%); subthreshold—138 (37%); while 180 (48%) did not score above threshold atany stage (never depressed).

2.3. Baseline profile and depression trajectories

The baseline profile of the sample is shown in Table 1 andis subdivided by depression trajectory category.

There was a significant difference in age among thedepression categories — those with persistent depressionhad the youngest average age. Those in the persistentdepression group were also less likely to have private healthinsurance, while those with subthreshold depression had thelowest prevalence of employment. There were no othermajor differences among the depression categories.

2.4. Predicting depression status longitudinally

The prevalence of depression was as follows during thefollow-up waves: 22% (75/335) at 3 months; 25% (75/302)at 6 months; 19% (60/312) at 12 months. The question ofwhether the vulnerabilities predicted depression statusduring follow-up is addressed in Table 2.

Unsurprisingly, depression status at baseline was a verystrong predictor of depression status during follow-up

Fig. 1. Depression trajectories, combining both depression scales.

227F. Doyle et al. / General Hospital Psychiatry 33 (2011) 224–231

(OR=36.7, 95% CI 14.2–94.5, Pb.001). As baselinedepression was also associated with vulnerabilities [13],we adjusted for baseline depression when assessing theassociation between individual vulnerabilities and subse-quent depression (Table 2A). Each vulnerability wassignificantly related to depression during follow-up, withORs ranging from 1.6 to 3.7.

Multivariate analysis (Table 2B), including each of thevulnerabilities and baseline depression in the model, showedthat depression status during follow-up was (at least

Table 1Sample description (baseline data plus at least one follow-up depression measurem

Total (N=375) Never depre(n=180)

DemographicsAge (years), mean (S.D.) 61.5 (10.5) 62.8 (9.9)Men 79% 82%Has a partner (1=yes) 75% 79%Employed (1=yes) 18% 22%Private health insurance 33% 38%Risk factor profileCurrent smoker 32% 27%Prior hypertension 48% 52%Prior diabetes 12% 13%Total cholesterol (mmol/l), mean (S.D.) (n=284) 4.6 (1.2) 4.7 (1.1)Prior coronary heart disease 29% 28%Prior revascularization 23% 20%HospitalizationThrombolysis 24% 28%Revascularization received 23% 25%Cardiac arrest confirmed 15% 17%Length of hospital stay, mean (S.D.) 8.6 (6.4) 8.4 (6.5)Left ventricular function (confirmed as b40%) 13% 12%Co-morbiditiesModified Charlson Co-morbidity Index score,median (interquartile range)

0 (0–1) 0 (0–0)

⁎ Pb.05.** Pb.01.*** Pb.001.

marginally) independently predicted by each of the vulner-abilities. Thus, although depression at baseline was thepredictor with the largest effect size of subsequentdepression, the effects of the theoretical vulnerabilitieswere not mediated by initial depression status.

2.5. Predicting depression trajectories

Table 3 shows the results of a multinomial logisticregression model predicting depression trajectories, withnever depressed as the reference category.

ent)

ssed Subthreshold depression(n=138)

Persistent depression(n=57)

χ2 (F)statistic

P value

61.5 (10.8) 57.4 (10.4) F=5.86 .003⁎⁎

76% 79% 1.48 .47772% 70% 2.94 .23010% 25% 9.55 .008⁎⁎

30% 21% 5.96 .051

35% 40% 4.21 .12247% 40% 2.34 .31111% 9% 0.86 .6504.5 (1.2) 4.6 (1.3) F=1.32 .26927% 37% 2.87 .23824% 33% 4.31 .116

20% 23% 3.35 .18819% 28% 2.56 .27816% 11% 1.29 .5268.5 (5.4) 9.0 (8.5) F=0.16 .85215% 14% 0.88 .644

0 (0–1) 0 (0–1) F=0.04 .956

Table 2Random effects logistic regression models predicting person depression status over time

Odds ratio 95% CI P value

(A) Adjusting for baseline depressionLTE-Q (Stressful life events, n=323, observations=829) 1.6 1.04 2.4 .030*PES-SV (low reinforcement, n=327, observations=830) 2.6 1.6 4.2 b.001***BJW (nonbelief in a just world, n=375, observations=949) 1.9 1.3 2.8 .001**Type D personality (n=375, observations=949) 3.7 1.7 8.3 .001**

(B) Multivariate (n=295, 756 observations)LTE-Q (Stressful life events) 1.5 0.95 2.2 .084PES-SV (low reinforcement) 2.1 1.3 3.5 .002**BJW (nonbelief in a just world) 1.8 1.1 2.7 .014*Type D personality 2.2 0.87 5.4 .097Baseline depression 14.2 4.9 41.2 b.001***

Overall multivariate model: χ2=49.5, df=5, Pb.001.⁎ Pb.05.⁎⁎ Pb.01.⁎⁎⁎ Pb.001.

able 3ultinomial logistic regression model predicting depression trajectories,ith never depressed as reference group (n=295)

Relative risk ratio 95% CI P value

ever depressed (reference) – – – –

ubthreshold depressionLTE-Q (Stressful life events) 1.8 1.2 2.5 .001**

PES-SV (low reinforcement) 1.3 0.94 1.7 .124BJW (nonbelief in a just world) 1.9 1.4 2.6 b.001***

Type D personality 1.6 0.89 2.9 .140

ersistent depressionLTE-Q (stressful life events) 2.6† 1.7 4.1 b.001**

PES-SV (low reinforcement) 2.5† 1.6 4.0 b.001***

BJW (nonbelief in a just world) 2.3 1.5 3.5 b.001***† **

228 F. Doyle et al. / General Hospital Psychiatry 33 (2011) 224–231

Each vulnerability was a significant, independent predic-tor of persistent depression. For example, when compared tothose who were never depressed, those with persistentdepression were more than twice as likely to have reportedbeing of Type D disposition or reported having elevatedstressful life events or reduced pleasant events in the yearprior to the follow-up period, or not to have just-worldbeliefs. Adding age, employment or health insurance statusto the model, as these differentiated some the trajectorygroups, had little effect on the results (data not shown). Forsubthreshold depression, only just-world beliefs and stressfullife events differentiated between this category and thenever-depressed category.

Visual inspection of the effect sizes for the subthresholddepression category would suggest that these were consis-tently smaller than those for predicting persistent depression.We tested whether the effect sizes for persistent depressionwere significantly larger than the effect sizes for thesubthreshold category — i.e., whether vulnerabilities hadsignificantly stronger effects for persistent depressionoverall. Post hoc Wald statistics confirmed that, with theexception of BJW, the effect sizes for the vulnerabilitieswhen predicting persistent depression were significantlylarger than the effects when predicting the subthresholdcategory (Pb.05 in each case, data not shown). Thus,elevated levels of stressful life events, reduced pleasantactivities and Type D personality predicted persistentdepression to an even greater extent than they did forsubthreshold depression. This finding illustrates the power ofsuch vulnerabilities for predicting persistent depression inthis population.

Type D personality 3.6 1.6 8.5 .0032=105.6, df=8, Pb.001, pseudo R2=0.18.⁎ Pb.05.⁎⁎ Pb.01.⁎⁎⁎ Pb.001.† Significant difference in effect size between subthreshold and

ersistent categories (Wald test, Pb.05).

3. Discussion

We longitudinally examined whether theoretical vul-nerabilities for depression were independent predictors of

depression, and depression trajectories, over 1 yearfollow-up in patients with ACS. Results showed notonly that the vulnerabilities independently predicteddepression status over time, but also predicted thedifferent depression trajectories. Furthermore, vulnerabil-ities were especially important for persistent depression,being significantly stronger predictors of this categoryover the subthreshold category.

That depressive vulnerabilities predicted depressionstatus over the follow-up confirms and strengthens thefindings of previous cross-sectional reports [13,17]. Per-haps more importantly, however, was that these vulnera-bilities were independently predictive of post-dischargedepression when controlling for baseline depression. To ourknowledge, this is the first such finding in the literature.

TMw

N

S

P

χ

p

229F. Doyle et al. / General Hospital Psychiatry 33 (2011) 224–231

That the vulnerabilities were independent predictorsprobably reflects the heterogeneous nature of the etiologyof depression, and that the vulnerabilities represent distinctcausal theories (i.e., interpersonal, behavioral, cognitive,along with personality [13,14]).

When predicting trajectories of depression, persistentdepression was consistently predicted by the vulnerabilitiesin comparison to those who were not depressed. Further-more, with the exception of just-world beliefs, these effectswere significantly larger than the effect sizes whenpredicting the subthreshold depression category, althoughthis is post hoc analysis and needs to be interpreted withcaution. Thus, clinicians need to be especially cognisant ofpatients reporting such theoretical vulnerabilities post-ACS,to determine the probable evolution of depression and thelevel of intervention needed. Although some of thevulnerabilities were nonsignificant for predicting subthresh-old depression, this may be due to the somewhat lowerpower and the smaller effect sizes. Future research shouldaddress the question of whether these depression trajectoriesdiffer in response to intervention and whether interventionstargeting these vulnerabilities can enhance quality of life.

The present results support some previous findingsregarding a number of the above vulnerabilities. Cognitions,at least in the form of illness perceptions, have also beenassociated with new episodes of depression post-myocardialinfarction [15]. That personality predicts subsequent depres-sion in cardiac patients has been demonstrated previously[7,8,32]. Furthermore, Martens et al. [7] also showed thatType D personality was predictive of persistence of differentcategories of depression over time. In contrast to our results,other research showed that stressful life events were notassociated with depressive symptoms 1 year after myocar-dial infarction [16]. However, as depression was onlymeasured at two time points, these analyses modeledprevalence of depression and not depression trajectories aswas done here. Also, the authors used a combineddepression and anxiety score, rather than just depressivesymptoms, which may explain the disparity in findings.Although these studies consolidate the findings of ourresearch, the present findings add to the literature bymeasuring the vulnerabilities simultaneously.

The trajectories we found closely match those from onestudy [7], but not others [8,33]. There may be a number ofreasons for the disparities — the scales used or thenumber of time points during follow-up may explain thesediffering trajectories.

Our findings differ somewhat from some previousresearch in that we generally did not show significantassociations among demographic factors or coronary diseaseor treatment indices and depression development/trajectories[7,8,10,11]. However, such results are inconsistent, forexample, Spijkerman et al. [10] have shown that womenwere more likely to be depressed post myocardial infarction,whereas others have not [11,17]. Previous research usingsimilar depression scales as used here has not shown evidence

of sex effects either [6,34]. As regards disease indices,generally it is accepted that coronary disease and depressiondo not correlate, although there is some controversy over thestatus of the relationship between left ventricular function anddepression [10,13,35,36]. One reason for this could be thedepression scales used in previous research— the full-lengthBDI has multiple somatic symptoms, and scores on this scalemay be more readily contaminated by coronary diseasesymptoms, which should not be the case in this study. Giventhe inconsistency in the literature, it is perhaps unsurprisingthat depression trajectories were not associated withdemographic or disease indices in the present study. Thepresent results, along with previous findings [13], demon-strate that these vulnerabilities were more important fordepression trajectories than coronary disease indices ordemographic factors. Only age, employment and healthinsurance status were associated with depression trajectories.As such, it is important to stress that, although these variablesare readily available clinically, they appear to be much lessimportant for predicting depression than vulnerabilities.

Unfortunately, history of depression was unavailable.This may be crucial in determining the persistence orotherwise of the episodes recorded here, as previous researchhas shown the importance of depression history forpredicting in-hospital and post-discharge depression[7,10,37–39]. The unavailability of history of depressionalso means that the categories analyzed here does not directlymatch those used in other research [40,41]. Furthermore, it isunclear whether the vulnerabilities measured in this studywould continue to predict subsequent depressive symptomsonce history of depression was controlled for. It is probablethat the BJW-S is not a comprehensive measure of cognitivedistortions; however, it independently predicted depressionat baseline and also in the longitudinal analysis here. It isunclear whether different cognitive distortions would bebetter predictors of depression or of certain depressiontrajectories, and future research should address this. We hadlittle power to include disease indices and sociodemographicvariables. However, it is unlikely that these variablescontributed much as they did not discriminate in univariateanalyses [13]. The analysis only contains those whocompleted at least one follow-up measure, and this limitsthe generalizability of the findings. Missing data across timepoints could have led to misclassification of participants,e.g., participants could be considered never depressed if theywere not depressed at baseline or at 12 months, but hadmissing data at the 3- and 6-month follow-up points.Strengths of the present study include the longitudinaldesign, multiple vulnerability measures and the ability toadjust for depression at baseline when predicting subsequentdepression status. This rules out the possibility that thevulnerabilities predicted depression at baseline simply due torecall bias.

The findings herein are unique in that, for the first time,theoretical depressive vulnerabilities have been shown topredict depression post-ACS and different trajectories of

230 F. Doyle et al. / General Hospital Psychiatry 33 (2011) 224–231

depression also. Furthermore, that these vulnerabilities wereparticularly important predictors of persistent depressionhighlights the need for clinicians to be aware of patients withsuch psychosocial risk factors or characteristics. The recentCOPES trial showed that allowing patient preference fortreatment (psychotherapy or antidepressants) in a stepped-care model could enhance patient satisfaction with depres-sion treatment [42]. Future studies could address the questionof whether the self-reported vulnerabilities as outlined herecorrelate with patient preference for depression therapy, todetermine whether such findings have the potential toenhance patient satisfaction or the therapeutic relationship.

Acknowledgments

We thank research assistants Isobel Jeffares, CiaraO'Connor and Janet Singh, and participating hospitals,staff and patients.

References

[1] Thombs BD, Bass EB, Ford DE, Stewart KJ, Tsilidis KK, Patel U,et al. Prevalence of depression in survivors of acute myocardialinfarction. J Gen Intern Med 2006;21:30–8.

[2] Jacobi F, Rosi S, Faravelli C, Goodwin R, Arbabzadeh-Bouchez S,Lepine JP. The epidemiology of mood disorders. In: Griez EJL,Faravelli C, Nutt DJ, Zohar J, editors. Mood disorders: Clinicalmanagement and research issues. New York: John Wiley & Sons Ltd.;2005. p. 3–34.

[3] Doyle F, Conroy RM, McGee HM, Delaney M. Depressive symptomsin persons with acute coronary syndrome: specific symptom scales andprognosis. J Psychosom Res 2010;68:121–30.

[4] van Melle JP, de Jonge P, Spijkerman TA, Tijssen JG, Ormel J,van Veldhuisen DJ, et al. Prognostic association of depressionfollowing myocardial infarction with mortality and cardiovascularevents: a meta-analysis. Psychosom Med 2004;66:814–22.

[5] Nicholson A, Kuper H, Hemingway H. Depression as an aetiologic andprognostic factor in coronary heart disease: a meta-analysis of 6362events among 146 538 participants in 54 observational studies. EurHeart J 2006;27:2763–74.

[6] McGee HM, Doyle F, Conroy RM, De La Harpe D, Shelley E. Impactof briefly-assessed depression on secondary prevention outcomes afteracute coronary syndrome: a one-year longitudinal survey. BMCHealthServ Res 2006;6:9.

[7] Martens EJ, Smith OR, Winter J, Denollet J, Pedersen SS. Cardiachistory, prior depression and personality predict course of depressivesymptoms after myocardial infarction. Psychol Med 2008;38:257–64.

[8] Kaptein KI, de Jonge P, van den Brink RH, Korf J. Course ofdepressive symptoms after myocardial infarction and cardiac progno-sis: a latent class analysis. Psychosom Med 2006;68:662–8.

[9] Doyle F, Conroy R, McGee H. Challenges in reducing depression-related mortality in cardiac populations: cognition, emotion, fatigue orpersonality? Health Psychol Rev 2007;1:137–72.

[10] Spijkerman TA, van den Brink RH, Jansen JH, Crijns HJ, Ormel J.Who is at risk of post-MI depressive symptoms? J Psychosom Res2005;58:425–32 [discussion 33–4].

[11] vanMelle JP, de Jonge P, KuyperAM,Honig A, ScheneAH, Crijns HJ,et al. Prediction of depressive disorder following myocardial infarctiondata from the Myocardial INfarction and Depression-Intervention Trial(MIND-IT). Int J Cardiol 2006;109:88–94.

[12] Doyle F, McGee HM, Conroy RM, Shelley E, De La Harpe D.Increase in observed mental health difficulties one year after acute

coronary syndrome: general practitioner survey. Ir J Med Sci 2007;176:205–9.

[13] Doyle F, McGee H, Conroy R, Delaney M. What predicts depressionin cardiac patients: Sociodemographic factors, disease severity ortheoretical vulnerabilities? Psychol Health, published online 29thOctober 2010, 10.1080/08870441003624398.

[14] Davidson KW, Rieckmann N, Lesperance F. Psychological theories ofdepression: potential application for the prevention of acute coronarysyndrome recurrence. Psychosom Med 2004;66:165–73.

[15] Dickens C,McGowan L, Percival C, Tomenson B, Cotter L, Heagerty A,et al. Negative illness perceptions are associated with new-onset depres-sion following myocardial infarction. Gen Hosp Psychiatry 2008;30:414–20.

[16] Dickens C, Percival C, McGowan L, Douglas J, Tomenson B, Cotter L,et al. The risk factors for depression in first myocardial infarctionpatients. Psychol Med 2004;34:1083–92.

[17] Rieckmann N, Burg MM, Gerin W, Chaplin WF, Clemow L, DavidsonKW. Depression vulnerabilities in patients with different levels ofdepressive symptoms after acute coronary syndromes. PsychotherPsychosom 2006;75:353–61.

[18] Wang H-Y, Chew G, Kung CT, Chung K-J, Lee WH. The use ofCharlson Comorbidity Index for patients revisiting the emergencydepartment within 72 hours. Chang Gung Med J 2007;30:437–44.

[19] Beck AT, Guth D, Steer RA, Ball R. Screening for major depressiondisorders in medical inpatients with the Beck Depression Inventory forPrimary Care. Behav Res Ther 1997;35:785–91.

[20] Beck AT, Steer RA, Ball R, Ciervo CA, Kabat M. Use of the BeckAnxiety and Depression Inventories for primary care with medicaloutpatients. Assess 1997;4:211–9.

[21] Scheinthal SM, Steer RA, Giffin L, Beck AT. Evaluating geriatricmedical outpatients with the Beck Depression Inventory-Fastscreen formedical patients. Aging Ment Health 2001;5:143–8.

[22] Golden J, Conroy RM, O'Dwyer AM. Reliability and validity of theHospital Anxiety and Depression Scale and the Beck DepressionInventory (Full and FastScreen scales) in detecting depression inpersons with hepatitis C. J Affect Disord 2007;100:265–9.

[23] Zigmond AS, Snaith RP. The Hospital Anxiety and Depression Scale.Acta Psychiatr Scand 1983;67:361–70.

[24] Bjelland I, Dahl AA, Haug TT, Neckelmann D. The validity of theHospital Anxiety and Depression Scale. An updated literaturereview. J Psychosom Res 2002;52:69–77.

[25] Brugha T, Bebbington P, Tennant C, Hurry J. The List of ThreateningExperiences: a subset of 12 life event categories with considerablelong-term contextual threat. Psychol Med 1985;15:189–94.

[26] Brugha TS, Cragg D. The List of Threatening Experiences: thereliability and validity of a brief life events questionnaire. ActaPsychiatr Scand 1990;82:77–81.

[27] Logsdon RG, Teri L. The Pleasant Events Schedule-AD: psychometricproperties and relationship to depression and cognition in Alzheimer'sdisease patients. Gerontology 1997;37:40–5.

[28] Furnham A. Belief in a just world: research progress over the pastdecade. Pers Individ Differ 2003;34:795–817.

[29] Lipkus IM, Dalbert C, Siegler IC. The importance of distinguishing thebelief in a just world for self versus for others: implications forpsychological well-being. Pers Soc Psychol Bull 1996;22:666–77.

[30] Denollet J. DS14: Standard assessment of negative affectivity,social inhibition, and Type D personality. Psychosom Med 2005;67:89–97.

[31] Jones BL, Nagin DS, Roeder K. A SAS procedure based on mixturemodels for estimating developmental trajectories. Sociol Methods Res2001;29:374–93.

[32] Pedersen SS, Ong AT, Sonnenschein K, Serruys PW, Erdman RA,van Domburg RT. Type D personality and diabetes predict the onsetof depressive symptoms in patients after percutaneous coronaryintervention. Am Heart J 2006;151:367.e1–6.

[33] Smith OR, Kupper N, Denollet J, de Jonge P. Vital exhaustionand cardiovascular prognosis in myocardial infarction and heart

231F. Doyle et al. / General Hospital Psychiatry 33 (2011) 224–231

failure: predictive power of different trajectories. Psychol Med 2011;41:731–8.

[34] Doyle F, McGee HM, De La Harpe D, Shelley E, Conroy RM. TheHospital Anxiety and Depression Scale depression subscale, but notthe BeckDepression Inventory-Fast Scale, identifies patients with acutecoronary syndrome at elevated risk of 1-year mortality. J PsychosomRes 2006;60:461–7.

[35] Carney RM, Freedland KE, Miller GE, Jaffe AS. Depression as a riskfactor for cardiac mortality and morbidity: a review of potentialmechanisms. J Psychosom Res 2002;53:897–902.

[36] van Melle JP, de Jonge P, Ormel J, Crijns HJ, van Veldhuisen DJ,Honig A, et al. Relationship between left ventricular dysfunction anddepression following myocardial infarction: data from the MIND-IT.Eur Heart J 2005;26:2650–6.

[37] Strik JJ, Lousberg R, Cheriex EC, Honig A. One year cumulativeincidence of depression following myocardial infarction and impact oncardiac outcome. J Psychosom Res 2004;56:59–66.

[38] Lesperance F, Frasure-Smith N, Talajic M. Major depression beforeand after myocardial infarction: its nature and consequences.Psychosom Med 1996;58:99–110.

[39] Sorensen C, Brandes A, Hendricks O, Thrane J, Friis-Hasche E,Haghfelt T, et al. Psychosocial predictors of depression in patients withacute coronary syndrome. Acta Psychiatr Scand 2005;111:116–24.

[40] Parker GB,Hilton TM,WalshWF,OwenCA,HerucGA,OlleyA, et al.Timing is everything: The onset of depression and acute coronarysyndrome outcome. Biol Psychiatry 2008;64:660–6.

[41] de Jonge P, van den Brink RH, Spijkerman TA, Ormel J. Only incidentdepressive episodes after myocardial infarction are associated with newcardiovascular events. J Am Coll Cardiol 2006;48:2204–8.

[42] Davidson KW, Rieckmann N, Clemow L, Schwartz JE, Shimbo D,Medina V, et al. Enhanced depression care for patients with acutecoronary syndrome and persistent depressive symptoms: coronarypsychosocial evaluation studies randomized controlled trial. ArchIntern Med 2010;170:600–8.