which factors predict the persistence of dsm-iv depression, anxiety, and somatoform disorders in the...

Journal of Psychosomatic Research 68 (2010) 21–28

Which factors predict the persistence of DSM-IV depression, anxiety, andsomatoform disorders in the medically ill three months post

hospital discharge?☆

Maria McKenziea,⁎, David M. Clarkea, Dean P. McKenzieb, Graeme C. Smitha

aPsychological and Behavioural Medicine Unit, Monash University School of Psychiatry, Psychology, and Psychological Medicine, Monash Medical Centre,Clayton, Australia

bDepartment of Epidemiology and Preventive Medicine, Monash University School of Public Health and Preventive Medicine, Melbourne, Australia

Received 25 July 2008; received in revised form 10 August 2009; accepted 11 August 2009

Abstract

Objective: This study sought to assess the persistence ofDSM-IV depression, anxiety, and somatoform disorders in asample of 206 medical patients 3 months after hospital dischargeand to examine which baseline factors predicted the persistence ofdisorder. Methods: Patients were interviewed using the MonashInterview for Liaison Psychiatry (a structured psychiatric interviewfor the medically ill) during admission and again at 3 monthspost discharge. Scales completed during admission elicitedsociodemographic data, psychiatric history, mental and physicalfunctioning, illness behavior, coping modes, and number of closerelationships. Best-subset logistic regression was employed to findthe best combination of these potential predictors of thepersistence of psychiatric disorder. Results: Persistence of anxietydisorders [n=43; 50.6%; 95% CI=39.5–61.6], depression (n=55;44.4%; 95% CI=35.4–53.5), and somatoform disorders (n=35;42.2%; 95% CI=31.3–53.0) was moderately high, with no

☆ This study was conducted in the Psychological and BehaviouralMedicine Unit, Monash University School of Psychiatry, Psychology, andPsychological Medicine, Monash Medical Centre, Clayton, Australia.

⁎ Corresponding author. Psychological and Behavioural Medicine Unit,Monash Medical Centre, 246 Clayton Road, Clayton, Victoria 3168,Australia. Tel.: +61 3 9594 1473; fax: +61 3 9594 6499.

E-mail address: [email protected] (M. McKenzie).

0022-3999/09/$ – see front matter © 2010 Elsevier Inc. All rights reserved.doi:10.1016/j.jpsychores.2009.08.004

statistically significant difference in the rate of persistence of thethree groups of disorder. Family psychiatric history, education, andpoorer physical and mental functioning during hospitalizationpredicted persistence of depression. Poorer mental functioning,less denial, and greater number of close relationships predictedpersistence of anxiety disorders. Higher levels of education, use ofacceptance–resignation as a coping mechanism, and greaterhypochondriasis predicted persistence of somatoform disorders.Conclusion: The belief that psychiatric disorders in hospitalizedmedically ill patients spontaneously remit after discharge is false.A substantial proportion persist for at least 3 months. Earlydetection and treatment is possible and warranted. Features of theillness (poorer physical and mental health) and personal and socialfactors identifiable at hospital admission identify patients at riskfor persistence.© 2010 Elsevier Inc. All rights reserved.

Keywords: Anxiety; Depression; Medical inpatients; Persistence; Prediction; Somatoform disorders

Introduction

Psychiatric morbidity, particularly depression and anxi-ety, is common in the hospitalized medically ill [1].

Prevalence rates range from 23% to 45% [2]. Furthermore,a psychiatric disorder comorbid with physical illness isassociated with low recovery rates, poor functioning andhealth-related quality of life, increased utilization of medicalservices and increased health costs, lost productivity, andgreater mortality [1,3–10]. Even mild depression and“subthreshold” disorders increase the risk of morbidity andmortality [11].

Despite this evidence, psychiatric disorders are often notrecognized or adequately treated in the general hospitalsetting [12–14]. Reasons for this include difficulties in

mailto:maria.mckenzie@mcri.�edu.au

http://dx.doi.org/10.1016/j.jpsychores.2009.08.004

22 M. McKenzie et al. / Journal of Psychosomatic Research 68 (2010) 21–28

applying diagnostic criteria [10,15]; a belief that depressionis not important, that it is normal in the medically ill, or that itwill resolve spontaneously [1,16,17]; pressure of time orphysician inexperience; concern about medication/diseaseinteractions; and uncertainty about the effectiveness oftreatment [16].

There have been relatively few studies of persistence ofpsychiatric disorder in the medically ill, and most haveinvolved small samples or short follow-up periods [16,18–20]. Physical illness, in particular chronic physical illness,and multiple medical conditions have been found to predictthe persistence of psychiatric disorder [20,21], as havesymptom perception, interpretation, and attribution [22].Poor physical functioning has been found to predictpersistent depressive disorder [23–25].

Other predictors of the persistence of psychiatric disorderidentified in the literature include severity of the mentaldisorder, poor mental functioning, demographic factors (ageand gender), social vulnerability (low-level education, lowersocioeconomic status, and living without a partner),psychobiological vulnerability (e.g., previous psychiatricillness and family psychiatric history), personality (avoidantor dependent), and sustaining factors (e.g., further negativelife events and lack of social support) [20,21,26,27].

The objectives of the present study were to examine thepersistence of the common forms of psychiatric illness(depression, anxiety, and somatoform disorders) 3 monthspost hospital discharge and to identify factors predictiveof persistence.

Method

The study was conducted at Monash Medical Centre, auniversity-affiliated general hospital in Melbourne, Austra-lia. All patients admitted to the general medical wards whowere able and willing to participate took part in a two-stageassessment procedure. All patients were screened with the36-item self-report General Health Questionnaire (GHQ)[28]. Those who scored above a 20/21 cutoff using binaryscoring (as per Goodchild and Duncan-Jones [29]) and whodid not meet the exclusion criteria of significant cognitiveimpairment, severe physical incapacity, or poor English wereadmitted to the study and, after having given their writteninformed consent, proceeded to diagnostic interview usingthe Monash Interview for Liaison Psychiatry (MILP) [30].Three hundred and twelve patients were interviewed andthese comprised the Time 1 sample.

Following the interview, the Global Assessment ofFunctioning (GAF) scale [31] and the Karnofsky rating ofphysical functioning [32] were completed as globalmeasures of mental and physical health, respectively, anda rating of the severity of illness episode was made. Thelatter was performed, in conjunction with the treatingphysician, using the Severity of Illness Index [33], whichrates stage of diagnosis, levels of complications and

dependency, nature of interventions, and response totherapy, and summarizes these data using a 4-point Likertscale ranging from 1 to 4. Following this, self-reportquestionnaires were administered.

After data entry, Diagnostic and Statistical Manual ofMental Disorders, Fourth Edition (DSM-IV) diagnoses wereobtained from the MILP data using a computerized scoringalgorithm [34]. The 312 Time 1 patients had a mean age of47.5 years (range, 18–85 years), and 190 (61%) were female.Two hundred and fifty-one (80.4%) of these patients had atleast one depressive, anxiety, or somatoform spectrumdisorder.

Face-to-face follow-up diagnostic interviews were con-ducted using the MILP interview approximately 3 monthsafter hospital discharge (Time 2) in the patient's home.Sixty-six percent (n=206) of the original sample wasavailable for follow-up. Reasons for nonparticipation atfollow-up were loss of contact (n=27; 13%), death (n=15;7%), too unwell (n=6; 3%), geographical inaccessibility(n=12; 6%), and refusal (n=46; 22%).

Self-report instruments administered during hospitalization

The Illness Behavior Questionnaire (IBQ) [35] is a 62-item questionnaire (in yes/no format) that assesses attitudes,ideas, affect, and attributions in relation to illness. It hasmostly been used in the study of somatoform disorders [36]and has been shown to distinguish between hypochondriacaland nonhypochondriacal patients [37] and between organicand nonorganic abdominal pain [38]. We were interested inthe subscales of general hypochondriasis (concerns abouthealth accompanied by a high level of anxiety), diseaseconviction (adamant belief in the presence of seriouspathology), and denial (refusal to admit the presence ofproblems in one's life).

The Medical Coping Modes Questionnaire (MCMQ) [39]is a 32-item measure of three illness-related copingstrategies: confrontation, avoidance, and acceptance–resig-nation. Confrontation has been shown to be used morefrequently by patients with acute life-threatening illness thanby patients with chronic illness [40]. Acceptance–resigna-tion, however, has been used by patients with littleexpectation of recovery. The use of avoidance or accep-tance–resignation in response to acute life-threateningsituations has been negatively associated with effectivecoping, while the use of acceptance–resignation in chron-ically ill patients has been associated with better coping [41].

The number of close relationships, as a measure ofsocial support, was assessed using one item of the Indicesof Social Functioning and Resources from the Health andDaily Living Form [42]. Low levels of social support havebeen associated with psychological distress and psychoso-matic complaints, and social resources have been found tomediate the effects of stressful life events on mood andphysical symptoms [43]. Lack of social support has beenidentified as a predictor of the persistence of psychiatricdisorder [20].

Table 1Persistence of depressive, anxiety, and somatoform disorders a

Time 1 diagnosisPersistence of diagnosisat Time 2 [n (%)] 95% CI

Depressive disorder b (n=124) 55 (44.4) 35.4–53.5Anxiety disorder c (n=85) 43 (50.6) 39.5–61.6Somatoform spectrum

disorder d (n=83)35 (42.2) 31.3–53.0

a Patients remaining within the broad diagnostic categories between

23M. McKenzie et al. / Journal of Psychosomatic Research 68 (2010) 21–28

Statistical analysis

For each of the DSM-IV categories, we selected thosepatients with a diagnosis in hospital (Time 1) andascertained the proportion of these patients with persistentdiagnoses at 3 months postdischarge (Time 2). In the case ofdepression, we took note of the full syndrome [majordepressive disorder (MDD)], as well as the minor categories,as persistent symptoms, even at a sub-syndromal level,produce significant ongoing morbidity [44,45]. Confidenceintervals for persistence were calculated using an exactprocedure for binary data [46].

It is statistically possible for confidence intervals forparameters such as persistence to overlap while still beingsignificantly different from each other [47]. The confidenceinterval of the difference between the two parameters istherefore the more relevant one [47]. A general methodknown as the bootstrap [48,49] was employed to generateconfidence intervals of the difference between proportionsusing software similar to that employed in recent psychiatricstudies [50,51]. Unlike conventional methods, the bootstrapallows for varying numbers of diagnoses for each person,such as anxiety and depression occurring together.

Logistic regression was used to identify predictors ofpersistence at Time 2 using variables at Time 1. Of theinformation available, we chose the most theoreticallymeaningful predictors in order to maximize the participant-to-variable ratio. These 13 predictors were age, sex,education, Karnofsky rating of physical functioning, pastpsychiatric history, family psychiatric history, baselineglobal assessment of functioning (GAF), general hypochon-driasis, denial and disease conviction (IBQ), coping mechan-isms, avoidance and acceptance–resignation (MCMQ), andnumber of close relationships.

Best-subset logistic regression [52,53] was then used tofind the best subset or combination of these predictors. Incontrast to stepwise logistic regression methods [47,48] thatadd or delete predictors one at a time, best-subset logisticregression identifies the best combination of variables fromall possible combinations. As recommended [52], the bestsubset overall was defined as that with the lowest value of Cp

[54], an index of model performance balanced against modelcost (number of predictors). Best-subset logistic regressionwas performed using XLMiner [55].

Otherwise, statistical analyses were conducted usingSPSS version 14.0 [56].

Time 1 and Time 2, although they may not have had the same specificdisorder at Time 2 that they had at Time 1.

b Depressive disorder includes MDD, MDD in partial remission,dysthymia, adjustment disorder with depressed mood, and adjustmentdisorder with mixed anxiety and depressed mood.

c Anxiety disorder includes generalized anxiety disorder, phobicdisorder, obsessive–compulsive disorder, depersonalization disorder, acutestress disorder, posttraumatic stress disorder, and adjustment disorder withanxiety. There were no diagnoses of panic disorder at Time 1.

d Somatoform spectrum disorder includes somatization disorder,hypochondriasis, pain disorder, undifferentiated somatoform disorder,psychological factors affecting medical condition, and conversion disorder.

Results

Sample characteristics

Two hundred and six patients completed follow-upinterviews at a mean of 91.0 (S.D.=11.1) days after hospitaldischarge. A logistic regression analysis, using “followed up/not followed up” as the dependent variable, revealed no

statistically significant difference between patients followedup and those lost to follow up with respect to age, sex,country of birth, religion, education, occupation, GAF score,and Karnofsky score. However, there were statisticallysignificant differences in marital status [separated patientswere less likely to be followed up; odds ratio (OR)=0.16;95% confidence interval (95% CI)=0.03–0.77] and medicaldiagnosis [patients with diagnoses of the respiratory(OR=0.31; 95% CI=0.10–0.93), gastrointestinal (OR=0.27;95% CI=0.10–0.74), and musculoskeletal (OR=0.30; 95%CI=0.10–0.91) systems were less likely to be followed up].There was no association between death and medicaldiagnosis (χ2=6.87, df=5, NS).

Patients who completed follow-up had a mean age of 47years (range, 18–85 years), 124 (60%) were female, 148(72%) were born in Australia, 145 (70%) had completed 10or more years of education, 113 (55%) were married or in ade facto relationship, 66 (32%) were living with family, and91 (44%) were supported by a pension. These patients wereadmitted to hospital with the following medical conditions:cardiovascular, 47 (23%); respiratory, 30 (15%); gastroin-testinal, 30 (14%); neurological, 25 (12%); musculo-skeletal, 21 (10%); “other”, 53 (26%).

Persistence of psychiatric disorder

Table 1 shows that 55 patients (44.4%; 95% CI=35.4–53.5) had a persistent depressive disorder, while 43 (50.6%;95% CI=39.5–61.6) and 35 (42.2%; 95% CI=31.3–53.0)patients had a persistent anxiety and somatoform spectrumdisorder, respectively. While anxiety disorders appearedmore persistent overall than depressive and somatoformspectrum disorders, these differences were not statisticallysignificant according to the bootstrap procedure describedearlier (the difference between persistence of anxietydisorders and persistence of depressive disorders, 95% CI=

Table 2Persistence of DSM-IV depressive disorders a

Time 1 diagnosis

Time 2 diagnosis

MDD [n (%)] 95% CI

Minor depression b

(partial remission)[n (%)] 95% CI

No depression(full remission)[n (%)] 95% CI

MDD with melancholic features (n=22) c 8 (36.4) 17.2–59.3 5 (22.7) 7.8–45.4 9 (40.9) 20.7–63.6MDD (n=37) d 12 (32.4) 18.0–49.8 8 (21.6) 9.8–38.2 17 (45.9) 29.5–63.1Dysthymia (n=11) e 1 (9.1) 0.2–41.3 4 (36.4) 10.9–69.2 6 (54.5) 23.4–83.3Adjustment disorder (n=67) 4 (6.0) 1.7–14.6 21 (31.3) 20.6–43.8 42 (62.7) 50.0–74.2

a Patients remained within the broad diagnostic categories between Time 1 and Time 2, although they may not have had the same type of disorder at Time 2that they had at Time 1.

b Minor depression includes cases where there are two, three, or four symptoms of depression. This includes cases of MDD in partial remission, adjustmentdisorder with depressed mood, adjustment disorder with mixed anxiety and depressed mood, and dysthymia.

c MDD with melancholic features at Time 2 is included with MDD.d Cases of MDD in partial remission at Time 1 (n=9) are not included.e There were no cases of “double depression” (MDD and dysthymic disorder occurring together) at Time 1.


−20.4 to 4.2; the difference between persistence of anxietydisorders and persistence of somatoform disorders, 95% CI=−6.3 to 19.7).

Table 2 illustrates the persistence of depressive disordersin more detail. We included the so-called “minor” or“subthreshold” disorder groups, yielding three syndromesequivalent to persistent MDD, minor depression includingMDD in partial remission, and full remission. The remissionrate of adjustment disorder was almost 63% (95% CI=50.0–74.2), with 42 of the 67 patients with no depression at Time2. Nine of 22 patients with MDD with melancholic features(almost 41%; 95% CI=20.7–63.6) recovered. Twenty of the37 patients (54.1%; 95% CI=36.9–70.5) with MDD while inhospital had persistent symptoms 3 months post hospitaldischarge. Dysthymic disorder endured in 5 of 11 patients(45.5%; 95% CI=16.7–76.6), but even adjustment disorderhad a persistence rate of approximately 37% (25 of 67patients; 95% CI=25.8–50.0) at 3 months.

Table 3 highlights the findings for anxiety disorders. Allcases of posttraumatic stress disorder had remitted by Time2; phobic disorders had the highest rate of persistence.

Table 3Persistence of DSM-IV anxiety disordersa,b,c


Generalized anxiety disorder (n=37) 7 (18.9) 8.0–35.2Phobic disorder (n=34) 15 (44.1) 27.2–62.1Obsessive–compulsive disorder(n=12)

4 (33.3) 9.9–65.1

Depersonalization disorder (n=14) 3 (21.4) 4.7–50.8Posttraumatic stress disorder (n=5) 0 0–52.2Adjustment disorder with anxiety(n=18)

1 (5.6) 0.1–27.3

a There were no cases of acute stress disorder at Time 1.b Diagnoses at Time 1 are not mutually exclusive: thirty-one patients

had more than one anxiety disorder.c Patients remained within the broad diagnostic category of anxiety

disorders between Time 1 and Time 2, although they may not have had thesame type of disorder at Time 2 that they had at Time 1.

Table 4 displays the persistence of DSM-IV somatoformspectrum disorders. Specific disorders within this groupgenerally had low persistence rates. Pain disorder had thehighest rate of persistence, with 5 of 20 patients (25%; 95%CI=8.7–49.1) with enduring disorder, while all cases ofconversion disorder had remitted.

Predictors of persistence of psychiatric disorder

Table 5 illustrates the results of the best-subset logisticregression for the three categories of DSM-IV diagnosis.Patients with a lower baseline GAF score and those with agreater number of years of education were significantlymore likely to have persistent depressive disorder 3 monthspost hospitalization. Patients with a family psychiatrichistory were twice as likely as patients without a familypsychiatric history to have a persistent depressive disorder.A lower Karnofsky score also predicted persistence ofdepressive disorders.

A lower baseline GAF score significantly predicted thepersistence of anxiety disorders 3 months post hospitaldischarge. A lower denial score (IBQ) and a greater

Table 4Persistence of DSM-IV somatoform spectrum disordersa,b


Somatization disorder (n=8) 1 (12.5) 0.3–52.7Hypochondriacal disorder (n=8) 1 (12.5) 0.3–52.7Conversion disorder (n=12) 0 0–26.5Pain disorder (n=20) 5 (25) 8.7–49.1Undifferentiated somatoform

disorder (n=25)4 (16) 4.5–36.1

Psychological factors affectingmedical condition (n=13)

1 (7.7) 0.2–36.0

a Diagnoses at Time 1 are not mutually exclusive: three patients hadmore than one somatoform disorder.

b Patients remained within the broad diagnostic category of somatoformspectrum disorders between Time 1 and Time 2, although they may not havehad the same type of disorder at Time 2 that they had at Time 1.

Table 5Predictors of the persistence of DSM-IV diagnoses 3 months post hospitaldischarge, determined by best-subset logistic regression

Persistent diagnosispredictor OR Significance 95% CI

Depressive disorders (n=113)GAF score (Time 1) ⁎ 0.96 0.037 0.93–1.00Family psychiatric history 1.99 0.101 0.87–4.55Education (years) ⁎⁎ 1.23 0.009 1.05–1.43Karnofsky score 0.97 0.096 0.93–1.01

Anxiety disorders (n=77)GAF score (Time 1) ⁎⁎⁎ 0.88 0.0001 0.82–0.94Denial score 0.73 0.059 0.52–1.01Number of close relationships 1.03 0.090 1.00–1.05

Somatoform spectrum disorders (n=76)Education (years) ⁎⁎ 1.41 0.001 1.14–1.74General hypochondriasis score ⁎ 1.34 0.036 1.02–1.76Acceptance–resignation score ⁎ 1.31 0.020 1.04–1.64

⁎ Pb.05.⁎⁎ Pb.01.⁎⁎⁎ Pb.001.


number of close relationships also predicted persistence ofanxiety disorders.

A greater number of years of education, a higheracceptance–resignation score (MCMQ), and a highergeneral hypochondriasis score (IBQ) significantly predictedpersistence of somatoform spectrum disorders.

Best-subset logistic regressions were also performed forthe specific diagnoses comprising the depressive, anxiety,and somatoform spectrum disorders. No clearly discernable“best” subset (as indicated by Cp value) or more clinicallymeaningful subsets were obtained; in addition, the samplesize for many specific disorders was small. These results,therefore, are not reported.

Discussion

This large follow-up study of psychiatric disorder in themedically ill has uniquely examined the contribution ofcoping style, illness behavior, hypochondriasis, and socialsupport to the persistence of depression, anxiety, andsomatoform disorders.

The rates of persistence of depression and anxietydisorders found here are similar to those reported by otherstudies of the medically ill [57–59], which range from 36%to 52% for depression and 49% for anxiety disorders [60].Higher rates of persistence of somatoform disorders havebeen reported elsewhere [22] and for major and minordepression in medically ill [27,61–63] and psychiatrically ill[9] samples.

Anxiety disorders in the current study were moreenduring (over 50%) than depression or somatoformdisorders; however, differences were not statistically signif-icant. Phobic disorders made the strongest contribution tothis persistence; they are, after all, generally chronic

disorders. On the other hand, adjustment disorders withanxiety and generalized anxiety disorders tended to remit.

The surprisingly high rate and persistence of somatoformspectrum disorders are of concern considering their contri-bution to health care demand and the lack of attention givento them in the health care system and in research [64,65].

With respect to persistence of depression, Koenig andGeorge [21] found, as we did, that family psychiatric historypredicted a lower remission of depression. Consistent withour findings, Smalbrugge et al. [66] found that a higher levelof education predicted persistence of depressive disorder,while Gilchrist and Gunn [67] found that a lowereducational level predicted persistence of depressive disor-der. A higher educational level also predicted persistence ofsomatoform disorders in the present study. Higher educationhas been found to be associated with lower prevalence ratesof mental disorders [68]. It may be, therefore, that peoplewith higher education are less at risk for developingpsychiatric disorders, but more at risk for persistence ifthey do develop them.

A lower Karnofsky rating of physical functioning hasgenerally been found to predict persistent depressivedisorder [27,58,59,62,69], although not in all studies [70].Physical illness is a strong predictor of depression [71], andchronic physical illness has been found to predictpersistence of depressive disorder [20,21]. It is likely thatif poor physical health and functioning persists ordeteriorates, it will be difficult to overcome depression.These results underscore the inextricable link betweenphysical and mental disorders and the essence of physical–psychiatric comorbidity.

A low baseline GAF score was a predictor of thepersistence of both depression and anxiety. Other studieshave replicated this finding in relation to depression[68,72,73].

Consistent with some studies of the medically ill [74], butinconsistent with others [45,75], our results indicated that alower denial score predicted persistence of anxiety disorders.

Interestingly, a greater number of close relationships atbaseline predicted persistence of anxiety disorders. Peoplewith anxiety disorders may exhibit dependent personalitytraits; therefore, a greater number of close relationshipscould result from such dependency. The relationships ofpersistently distressed patients have been found to worsenover time [24], and poor perceived social support isassociated with persistent anxiety [70,76]. It is possible thatthe number of relationships does not equate to theperceived quality of relationships [77]. Alternatively,although those with a greater number of close relationshipsare generally less prone to developing mental disorders[68], their mental disorders may be more persistent if theydo develop them.

A greater number of years of education and higher generalhypochondriasis and acceptance–resignation scores com-prised the best subset of factors predicting persistence ofsomatoform disorders. Somatizing patients have a high


degree of hypochondriasis [78] and are more likely to makeillness attributions with somatic symptoms, with more illnessbehavior manifested by body scanning and seeking diag-nostic examinations [79]. This may result in greater fear andworry about health and greater utilization of services.Patients with hypochondriasis and somatization disorderbelieve that they are weak and have a “catastrophizing”interpretation of bodily complaints [78]. Consequently, theymay experience hopelessness, helplessness, and a greaterinclination to use acceptance–resignation as a copingmechanism. Acceptance–resignation is negatively linked toeffective coping [41] and may contribute to the persistence oftheir somatoform disorder.

Limitations of the present study include, first, incom-plete follow-up data. It is possible that persistingpsychiatric illness may influence patients' willingness toparticipate at Time 2. Although we cannot ascertainwhether those not followed up had a persistent psychiatriccondition, we do know that there were few statisticallysignificant differences between patients followed up andthose not followed up, including on the GAF score. Despitethis attrition, the Time 2 sample of 206 patients, repre-senting 66% of the Time 1 sample, was larger than thesamples of most studies of the medically ill. We were notable to discern systematic bias that would be expected toaffect the results. In addition, a reliable method of diagnosisincorporating standard criteria was used, further increasingthe validity of our methodology.

Second, we did not collect data on medical conditionat follow-up. It remains unclear therefore whetherpsychiatric disorders persisted because the medicalcondition was unresolved or because the psychiatricdisorder became autonomous.

Third, where we looked at persistence for collapsedcategories of disorder rather than for each specific disorder,rates of persistence will be inflated in comparison to those forthe restricted categories. (For instance, compare persistencerates for the general categories in Table 1 with those for thespecific categories in Tables 2–4.) A specific somatoformdisorder in a patient at Time 1, for instance, will be consideredto have persisted if any somatoform disorder is present atTime 2 in the analysis of the collapsed category of somato-form disorders.

Last, the high level of somatoform spectrum disordersreported here may be due to the inclusion of undifferentiatedsomatoform disorder in the structured interview anddiagnostic algorithm. This category is often not used inresearch and may be considered a “subsyndromal” condition.Its inclusion, however, highlights the frequency andpersistence of unexplained physical symptoms.

Findings from the present study, which identify riskfactors for persistence of depression, anxiety, and somato-form disorders from data obtained during the initial patientassessment, should facilitate both the early recognition ofpatients at risk, and potentially, early intervention to reduceongoing morbidity.

Acknowledgments

This work was supported by a grant from the NationalHealth and Medical Research Council of Australia.

References

[1] Cassem E. Depressive disorders in the medically ill: an overview.Psychosomatics 1995;36:S2–S10.

[2] Clarke DM, Kissane DW, Trauer T, Smith GC. Demoralization,anhedonia and grief in patients with severe physical illness. WorldPsychiatry 2005;4:96–105.

[3] Cole MG. Does depression in older medical inpatients predictmortality? A systematic review.GenHosp Psychiatry 2007;29:425–30.

[4] Cooper J, Harris Y, McGready J. Sadness predicts death in olderpeople. J Aging Health 2002;14:509–26.

[5] Creed F, Morgan R, Fiddler M, Marshall S, Guthrie E, House A.Depression and anxiety impair health-related quality of life and areassociated with increased costs in general medical inpatients.Psychosomatics 2002;43:302–9.

[6] Egede LE. Major depression in individuals with chronic medicaldisorders: prevalence, correlates and association with health resourceutilization, lost productivity and functional disability. Gen HospPsychiatry 2007;29:409–16.

[7] Furlanetto L, von Ammon Cavanaugh S, Bueno J, Creech S, Powell L.Association between depressive symptoms and mortality in medicalinpatients. Psychosomatics 2000;41:426–32.

[8] Hansen M, Fink P, Frydenberg M. Follow-up on mental illness inmedical inpatients: health care use and self-rated health and physicalfitness. Psychosomatics 2004;45:302–10.

[9] Keitner G, Ryan C, Miller I, Kohn R, Epstein N. 12-Month outcome ofpatients with major depression and comorbid psychiatric or medicalillness (compound depression). Am J Psychiatry 1991;148:345–50.

[10] von Ammon Cavanaugh S, Furlanetto LM, Creech SD, Powell LH.Medical illness, past depression, and present depression: a predictivetriad for in-hospital mortality. Am J Psychiatry 2001;158:43–8.

[11] Glassman A. Does treating post-myocardial infarction depressionreduce medical mortality? Arch Gen Psychiatry 2005;62:711–2.

[12] Koenig H. Physician attitudes toward treatment of depression in oldermedical inpatients. Aging Ment Health 2007;11:197–204.

[13] Rapp S, Parisi S, Walsh W, Wallace C. Detecting depression in elderlymedical inpatients. J Consult Clin Psychol 1988;56:509–13.

[14] Wancata J, Windhaber J, Bach M, Meise U. Recognition of psychiatricdisorders in nonpsychiatric hospital wards. J Psychosom Res 2000;48:149–55.

[15] Koenig H, George L, Peterson B, Pieper C. Depression in medically illhospitalized older adults: prevalence, characteristics, and course ofsymptoms according to six diagnostic schemes. Am J Psychiatry 1997;154:1376–83.

[16] Koenig H. Recognition of depression in medical patients with heartfailure. Psychosomatics 2007;48:338–47.

[17] Rapp S, Davis K. Geriatric depression: physicians' knowledge,perceptions, and diagnostic practices. Gerontologist 1989;29:252–7.

[18] Furukawa T, Kiturama T, Takahashi K. Time to recovery of aninception cohort with hitherto untreated unipolar depressive episodes.Br J Psychiatry 2000;177:331–5.

[19] Judd J. The clinical course of unipolar major depressive disorders.Commentary. Arch Gen Psychiatry 1997;54:989–91.

[20] Spijker J, de Graaf R, Bijl R, Beekman A, Ormel J, Nolan W.Determinants of persistence of major depressive episodes in thegeneral population. Results from the Netherlands Mental HealthSurvey and Incidence Study (NEMESIS). J Affect Disord 2004;81:231–40.

[21] Koenig H, George L. Depression and physical disability outcomes indepressed medically ill hospitalized older adults. Am J GeriatrPsychiatry 1998;6:230–47.


[22] Bailer J, Witthoft M, Bayerl C, Rist F. Syndrome stability andpsychological predictors of symptom severity in idiopathic andenvironmental intolerance and somatoform disorders. Psychol Med2007;32:271–81.

[23] Derks W, De Leeuw R, Winnubst J, Hordijk GJ. Elderly patients withhead and neck cancer: physical, social and psychological aspects after1 year. Acta Otolaryngol 2004;124:509–14.

[24] Dew MA, Myaskovsky L, Switzer GE, DiMartini AF, Schulberg HC,Kormos RL. Profiles and predictors of the course of psychologicaldistress across four years after heart transplantation. Psychol Med2005;35:1215–27.

[25] Hopwood P, Stephens R. Depression in patients with lung cancer:prevalence and risk factors derived from quality-of-life data. J ClinOncol 2000;18:893–903.

[26] Massion AO, Dyck IR, Shea MT, Phillips KA, Warshaw MG, KellerMB. Personality disorders and time to remission in generalised anxietydisorder, social phobia, and panic disorder. Arch Gen Psychiatry 2002;59:434–40.

[27] Schrader G, Cheok F, Hordacre A-L, Guiver N. Predictors ofdepression three months after cardiac hospitalization. PsychosomMed 2004;66:514–20.

[28] Goldberg DP, Williams P. A user's guide to the General HealthQuestionnaire. Windsor, England: NFER-Nelson, 1988.

[29] Goodchild ME, Duncan-Jones P. Chronicity and the General HealthQuestionnaire. Br J Psychiatry 1985;146:55–61.

[30] Clarke DM, Smith GC, Herrman HE. A comparative study ofscreening instruments for mental disorders in general hospital patients.Int J Psychiatry Med 1993;23:323–37.

[31] Endicott J, Spitzer R, Fleiss J, Cohen J. The Global Assessment Scale:a procedure for measuring overall severity of psychiatric disturbance.Arch Gen Psychiatry 1976;33:766–71.

[32] Schag C, Heinrich R, Ganz P. Karnofsky performance statusrevisited: reliability, validity and guidelines. J Clin Oncol 1984;2:187–93.

[33] Horn SD, Horn RA. Reliability and validity of the Severity of IllnessIndex. Med Care 1986;24:159–78.

[34] Yap R, Clarke D. An expert system for psychiatric diagnosis using theDSM-III-R, DSM-IV and ICD-10 classifications. Proc AMIA 1996:229–33.

[35] Pilowsky I, Spence N. Patterns of illness behaviour in patients withintractable pain. J Psychosom Res 1975;19:279–87.

[36] Pilowsky I, Spence N, Cobb J, Katsikitis M. The Illness BehaviourQuestionnaire as an aid to clinical assessment. Gen Hosp Psychiatry1984;6:123–30.

[37] Pilowsky I. Dimensions of hypochondriasis. Br J Psychiatry 1967;113:89–93.

[38] Joyce PR, Bushnell JA, Walshe JWB, Morton JB. Abnormal illnessbehaviour and anxiety in acute non-organic abdominal pain. Br JPsychiatry 1986;149:57–62.

[39] Feifel H, Strack S, Nagy V. Degree of life threat and differential use ofcoping modes. J Psychosom Res 1987;31:91–9.

[40] Lipowski Z. Physical illness, the individual and the coping processes.Psychiatry Med 1970;1:91–102.

[41] Feifel H, Strack S, Nagy V. Coping strategies and associated featuresof medical inpatients. Psychosom Med 1987;49:616–25.

[42] Moos R, Cronkite R, Billings A, Finney J. Health and Daily LivingForm manual. Palo Alto: Social Ecology Laboratory, Department ofPsychiatry and Behavioral Sciences, Stanford University and theVeterans Administration Medical Center, 1984.

[43] Billings A, Moos R. The role of coping responses and socialresources in attenuating the stress of life events. J Behav Med 1981;4:139–57.

[44] Beck D, Koenig H. Minor depression: a review of the literature. Int JPsychiatry Med 1996;26:177–209.

[45] Pincus H, Davis W, McQueen L. ‘Subthreshold’ mental disorders: areview and synthesis of studies of minor depression and other ‘brandnames’. Br J Psychiatry 1999;174:288–96.

[46] McKenzie D, Vida S, Mackinnon A, Onghena P, Clarke D. Accurateconfidence intervals for measures of test performance. Psychiatry Res1997;69:207–9.

[47] Wolfe R, Hanley J. If we're so different, who do we keep overlapping?Can Med Assoc J 2002;166:65–6.

[48] Armitage P, Berry G, Matthews JNS. Statistical methods in medicalresearch. 4th ed. Oxford, UK: Blackwell, 2002.

[49] Efron B, Tibshirani RJ. An introduction to the bootstrap. New York:Chapman and Hall, 1993.

[50] McFarlane AC, McKenzie DP, Van Hooff M, Browne D. Somatic andpsychological dimensions of screening for psychiatric morbidity: acommunity validation of the SPHERE Questionnaire. J PsychosomRes 2008;65:337–43.

[51] McKenzie DP, Mackinnon AJ, Peladeau N, Bruce PC, Onghena P,Clarke DM, et al. Comparing correlated kappas by resampling: is onelevel of agreement significantly different from another? J Psychiatr Res1996;30:483–92.

[52] Hosmer DW, Lemeshow S. Applied logistic regression. 2nd ed. NewYork: Wiley, 2000.

[53] King JE. Running a best-subsets logistic regression: an alternative tostepwise methods. Educ Psychol Meas 2003;63:392–403.

[54] Mallows CL. Some comments on Cp. Technometrics 1973;15:661–75.[55] CYTEL Software Corporation. XLMiner, version 3. Cambridge

(Mass): CYTEL Software Corporation, 2007.[56] SPSS, Inc.. SPSS for Windows, version 14.0. Chicago (Ill): SPSS, Inc.,

2005 [computer software].[57] Mayou R, Hawton K, Feldman E. What happens to medical patients

with psychiatric disorder? J Psychosom Res 1988;32:541–9.[58] Koenig HG. Predictors of depression outcomes in medical inpatients with

chronic pulmonary disease. Am J Geriatr Psychiatry 2006;14:939–48.[59] Koenig HG. Depression outcome in inpatients with congestive heart

failure. Arch Intern Med 2006;166:991–6.[60] Lam WWT, Chan M, Wai Ka H, Fielding R. Treatment decision

difficulties and post-operative distress predict persistence of psycho-logical morbidity in Chinese women following breast cancer surgery.Psycho-oncology 2007;16:904–12.

[61] Cole M, Bellavance F. Depression in elderly medical inpatients: ameta-analysis of outcomes. Can Med Assoc J 1997;157:1055–60.

[62] Cole MG, McCusker J, Ciampi A, Windholz S, Latimer E, Belzile E.The prognosis of major and minor depression in older medicalinpatients. Am J Geriatr Psychiatry 2006;14:966–75.

[63] Fenton FR, Cole MG, Engelsmann F, Mansouri I. Depression in oldermedical patients: one year course and outcome. Int J Geriatr Psychiatry1997;12:389–94.

[64] Creed F. Should general psychiatry ignore somatization andhypochondriasis? World Psychiatry 2006;5:146–50.

[65] Barsky A, Orav E, Bates D. Somatization increases medical utilizationand costs independent of psychiatric and medical comorbidity. ArchGen Psychiatry 2005;62:903–10.

[66] Smalbrugge M, Jongenelis L, Pot AM, Eefsting JA, Ribbe MW,Beekman AT. Incidence and outcome of depressive symptoms innursing home patients in the Netherlands. Am J Geriatr Psychiatry2006;14:1069–76.

[67] Gilchrist G, Gunn J. Observational studies of depression in primarycare: what do we know? BMC Fam Pract 2007;11:8–28.

[68] Stansfield S, Rasul F. Psychosocial factors, depression and illness. In:Steptoe A, editor. Depression and physical illness. Cambridge:Cambridge University Press, 2007. p. 19–49.

[69] Schleifer S, Macari-Hinson M, Coyle D, Slater W, Kahn M, Gorlin R,et al. The nature and course of depression following myocardialinfarction. Arch Intern Med 1989;149:1785–9.

[70] Hipkins J, Whitworth M, Tarrier N, Jayson G. Social support, anxietyand depression after chemotherapy for ovarian cancer: a prospectivestudy. Br J Health Psychol 2004;9:569–81.

[71] Wilhelm K, Mitchell P, Slade T, Brownhill S, Andrews G. Prevalenceand correlates of DSM-IV major depression in an Australian nationalsurvey. J Affect Disord 2003;75:155–62.


[72] Ezquiaga E, Garcia A, Bravo F, Pallares T. Factors associated withoutcome in major depression: a 6-month prospective study. SocPsychiatry Psychiatr Epidemiol 1998;33:552–7.

[73] Picardi A, Porcelli P, Mazzotti E, Fassone G, Lega I, Ramieri L, SagoniE, Pasquini P. Alexithymia and global psychosocial functioning: astudy on patients with skin disease. J Psychosom Res 2007;62:223–9.

[74] Folks D, Freeman A, Sokol R, Thurstin A. Denial: predictor ofoutcome following coronary bypass surgery. Int J Psychiatry Med1988;18:57–66.

[75] Classen C, Koopman C, Angell K, Spiegel D. Coping styles associatedwith psychological adjustment to advanced breast cancer. HealthPsychol 1996;15:434–7.

[76] Glynn S, Shetty V, Elliot-Brown K, Leathers R, Belin T, Wang J.Chronic posttraumatic stress disorder after facial injury: a 1-yearprospective cohort study. J Trauma 2007;62:410–8.

[77] Pedersen S, Middel B, Larsen M. The role of personality variables andsocial support in distress and perceived health in patients followingmyocardial infarction. J Psychosom Res 2002;53:1171–5.

[78] Rief W, Hiller W, Margraf J. Cognitive aspects of hypochondriasisand the somatization syndrome. J Abnorm Psychol 1998;107:587–95.

[79] Rief W, Nanke A, Emmerich J, Bender A, Zech T. Causal illnessattributions in somatoform disorders: associations with comorbidityand illness behavior. J Psychosom Res 2004;57:367–71.

which factors predict the persistence of dsm-iv depression, anxiety, and somatoform disorders in the...

Documents