10.1192/bjp.167.3.315Access the most recent version at DOI: 1995, 167:315-323.BJP 

D Quinton, L Gulliver and M Rutterdisorder and social functioning.A 15-20 year follow-up of adult psychiatric patients. Psychiatric

Referenceshttp://bjp.rcpsych.org/content/167/3/315#BIBLThis article cites 0 articles, 0 of which you can access for free at:

permissionsReprints/

permissions@rcpsych.ac.ukto To obtain reprints or permission to reproduce material from this paper, please write

to this article atYou can respond http://bjp.rcpsych.org/letters/submit/bjprcpsych;167/3/315

from Downloaded

The Royal College of PsychiatristsPublished by on November 28, 2014http://bjp.rcpsych.org/

http://bjp.rcpsych.org/site/subscriptions/ go to: The British Journal of PsychiatryTo subscribe to

British Journal of Psychiatry (1995), 167, 315—323

Long-term follow-up studies of psychiatric disordersare important for clinical practice and for advancesin nosology. For the clinician they can provideinformationon prognosisortestassumptionsaboutthe course of disorders that derive from clinicalobservation. For nosology they provide a meansof testing diagnostic distinctions on the basis ofconsistency in symptom patterns over time and ofdifferences in course and outcome (Kendell, 1989).

Studies of the outcome for adult psychiatricdisorders ten or more years after the first episode,satisfying adequate sampling requirements and withsystematic diagnoses, are still rare. In addition, formost studies, problems inevitably arise throughchanges in diagnostic schemes over time, andcomparisons between them are further complicatedby variations in the initial severity of the disorderand on whether first and subsequent referrals aredistinguished.

A strong case has been made for including broaderfeatures of social functioning as well as measures ofepisodic disorder in assessing outcome (Shepherd eta!, 1989),but the reason for poorer social functioningis hard to assess without data on personality disorderin the original data sets. Since personality disorders,by definition, involve long-term difficulties in socialfunctioning, it is crucial to include some measure ofthem when considering the outcome of episodicillnesses (Duggan ci a!, 1991). Personality disorderis likely to be an important influence on later socialfunctioning, as well as being a risk factor forsubsequent episodes of illness through the life eventsor difficulties it promotes.

Previousresearch

Episodic or ‘¿�mentalstate' disorders

We have found no studies that compare the longterm outcome of representative groups of patientswith different diagnoses, the great majority of studiesbeing restricted to single or related illnesses. Findingshave shown surprisingly little difference betweenbroad diagnostic groupings with regard to the recurrence of episodes of illness or the likelihood ofpoor subsequent social functioning. Follow-ups ofschizophrenic patients have shown remission ratesof 22—26%up to 22 years after first admission, andpoor social functioning in 25-35% of patients (Huberci a!, 1980; Biehl ci a!, 1986; Shepherd ci a!, 1989).

Studies of depressive disorders show strikinglysimilar levels of persistence. The poorest outcomeshave been reported by Lee & Murray (1988) in theirfollow-up of Maudsley depressive in-patients. Over18 years, almost all subjects relapsed at some timeto a severity satisfying Research Diagnostic Criteriafor disorder, although 25% saw themselves asessentially free of symptoms or with only minorrecurrence. At the other extreme, 34% experiencedchronic severe distress and marked social impairment. Similar figures have been reported for an inpatient sample in Sydney (Kiloh eta!, 1988;Andrewsci a!, 1990), but, whereas Lee & Murray's sampleshowed a high rate of other psychiatric problems inlater years, this was not so for the Sydney group.In both studies the prognosis became poorer aspsychotic symptoms worsened.

315

A 15—20Year Follow-Up of Adult Psychiatric PatientsPsychiatric Disorder and Social Functioning

DAVID QUINTON, LESLEY GULLIVER and MICHAEL RUTTER

Background.An exploratorystudywas undertakenof the importanceof personalitydisorderin predicting the long-term outcome for both episodic disorders and social functioning.Method.In 1966—67,a representativeseriesofpatientswithchildren,freeofepisodicillnessfor at leastoneyear,was sampledfromthe CamberwellPsychiatricRegisterandsystematicallyassessedovera four-yearperiod,usingmeasuresof knownreliabilityandvalidity.Psychiatricdisorderwas measuredusinga PSE-compatibleinstrument.The follow-upafter 15—20yearsusedthe PSEand a systematicassessmentof socialfunctioning.Results.Overalloutcomeswere similaracrossdiagnoses,but an initialcategoricaldiagnosisof personalitydisorderpredictedmuchpooreroutcomeson psychiatricand socialmeasuresfor patientswith unipolardepressivedisordersthan for those with other diagnoses.Conclusions. The findings indicate the importance for prognosis of including a systematicassessmentof personalitydisorderin the clinicalassessmentof patients with depressivedisorder.

316 QUINTON ET AL

We have found only one long-term study ofpatients with anxiety disorders. In a prospective studyover six to eight years of 40 patients with anxietydisorders, Kriegeta! (1987) found at follow-up overhalf with chronic and severe symptoms and only onethird free of marked symptoms. Sixty-seven per centhad severe difficulties across most social role areas.

Long-term natural history of personality disorders

As yet there are few long-term follow-up investigations of personality disorders using currentcriteria (Costa & McRae, 1986; Drake & Vaillant,1988). Moreover, predictions have usually been madeto continuity in the same disorder, narrowlydefined,rather than to a wider range of outcomes. Stone(1993) reviewed the follow-up data, especially on hisown series of patients with borderline personalitydisorder, over 10—30years. Two-fifths had continuing clinically ratable personality difficulties, andthe group as a whole showed poorer social functioningon interpersonal indicators such as stable cohabitations or family formation.

Episodic disorders and personality disorders

Recent studies have shown the substantial comorbidity between episodic and personality disorders(Vize & Tyrer, 1994). In clinic samples this may applyto half or more of patients (Rutter& Quinton, 1984;Friedman et a!, 1987). In a study of personalitydisorder in a sample that combined a generalpopulation group and the relatives of psychiatricpatients, Zimmerman & Coryell (1989) found thosewith a diagnosis of personality disorder showed arange of Axis I disorders as well, including majordepression in 3801oof cases.

Personality disorders and the outcomeof episodic disorders

Studies of the relevance of personality to theprognosis of episodic disorders have predominantlyexamined the importance of personality traits ratherthan of a categorical psychiatric diagnosis ofdisorder. The most frequently repeated finding onpersonality traits has been an association betweenhigh neuroticism or introversion scores, as measuredon the Eysenck Personality Inventory (EPI) or somerelated instrument, and poorer outcome in affectivedisorders (Weissman et a!, 1978; Chancy et a!,1981), even over 18 years (Duggan eta!, 1990). Oneearlier study did use a categorical clinical rating of“¿�abnormalpremorbid personality―. This was the4—6-yearfollow-up by Greer & Cawley (1966) of

Maudsley patients with neurotic disorders. Theyfound a much poorer outcome on symptoms andsocial functioning for patients with abnormal premorbid personalities.

The issue of whether the personality featuresassociated with episodic disorders are state dependent, a consequence of the episodic illness, adistinct aspect of functioning, or a reflection ofa common underlying pathology remains unresolved(Akiskal et a!, 1983; Docherty et a!, 1986). It seemsunlikely that there will be an answer to this questionthat will encompass all forms of episodic disorderor abnormal personality feature or type (Rutter,1987; Vize & Tyrer, 1994), but long-term prospectivestudies including both axes are an essential strategyfor approaching this issue.

None of the studies of episodic or personalitydisorder reviewed above focused on the relationshipbetween the two in predicting the outcome for either.In this paper we examine the psychiatric and socialoutcome after 15—20years for a representativesampleof adult in-and out-patientswithand withoutpersonality disorder, as defined on clinical criteriacurrent in 1966.

Sample

Original study

Method

In 1966—67an epidemiologically based representativesample of 137adult psychiatric patients was collectedusing the Camberwell Psychiatric Register (Wing &Halley, 1972)as part of a study of the effects of newparental mental disorder on children (Rutter &Quinton, 1984). This sample included a two-in-fiverandom sample of a consecutive seriesof in- and outpatients with children under the age of 15 years, whohad had no contact with psychiatric services for atleast one year, and for whom English was the mainlanguage spoken at home (n = 84). In order toincrease the numbers in key cells, this series wassupplemented with a further consecutive series of 53male patients and patients with psychoses. Bothseries of patients were ascertained in the same way.The total sample was thus representative within sexand diagnostic subgroups. However, since thesampling criteria were intended to omit those patientswith chronic disorder —¿�as indicated by regularhospital contacts —¿�the sample was representativeonly of those patients experiencing a new episode.

Clinical and social data were collected in separateinterviews with patients and spouses as soon aspossible after the new contact with services, and then

31715-20 YEAR FOLLOW-UP OF PSYCHIATRIC PATIENTS

at yearly intervals over the following four years, a totalof five assessment points (Rutter & Quinton, 1984).

Follow-up

The follow-up study of the patients was secondaryto a follow-up of their children, with the follow-uptaking place when the children were between the agesof 21 and 28.

In order that the children's follow-up should bebased on adequate data concerning their childhoodpsychiatric state, six cases with extensive missing datain the original study or where all children were tooyoung at that time for an assessment were omitted.Three of the remaining sample of 131 patients haddied during the course of the original study and afurther 12by the time of follow-up. Death certificateswere obtained on all of these. Information wasavailable on 98 of the 116 remaining patients. Thiscame from 83 direct interviews, 14 interviews withspouses or children, and the current case notes fromone patient with chronic illness. The mean age ofmale patients at follow-up was 53.9 (s.d. 7.5), andfemale patients 49.9 (s.d. 8.7). Eighteen patients wereuntraced or refused to be interviewed. Possible biasesin the studied sample were examined and no statistically significant differences were found on either clinicalor social variables between the patients for whomfollow-up data were available and those who werenot recontacted or had died. However, those withpoor marriages in the original study were somewhatmore likely to have refused or remained untraced.

Measures of disorder

The assessment measures were chosen to maximisecomparability of categorisations and criteria betweenthe original and follow-up studies. However, revisionof the original classifications was inevitable becauseof changes in diagnostic schemes and criteria.

Original study

Episodic disorders. Psychiatric diagnoses in theoriginal study were made on a systematic clinical interview covering 32 clinical symptoms and 14 symptomrelated behaviours as manifested in the family context(e.g. overt misery, social withdrawal, irritability),together with information on psychiatric history andsocial functioning from both the patient's and spouse'saccounts. This interview (the Family Illness Study(FIS) instrument) was developed in parallel with theearly versions of the Present State Examination (PSE;Wing ci a!, 1974) and shared similar interviewingmethods, symptoms definitions and cut-off points,although the PSE covered more symptoms.

Personality disorders. A diagnosis of personalitydisorder required evidenceof a pervasiveand persistentpattern of maladaptive interpersonal functioningfrom at least early adolescence not explicable in termsof any recognisable mental illness, and was originallydivided into antisocial, schizoid, depressive, anxious,obsessional and immature/inadequate categories.

Follow-up

Episodes of psychiatric disorder and the pattern ofsocial functioning were rated for the five years beforethe follow-up interview. The measures were as follows.

Episodic psychiatric disorders. An episode wasrated if the patient had contacted a psychiatrist eitheras an in-patient or an out-patient during the previousfive years, or if there was a period greater than twoweeks during which he or she reported social-roleimpairment (e.g. time off work, marital difficulties,reduction in social activities) associated with psychiatric symptoms. Because of its similarity to theFIS instrument, the PSE was used to make diagnosesof episodic disorders, using the CATEGOcomputerscoring system (Wing ci a!, 1974). The pattern ofillness was rated as ‘¿�episodic'if there was at leastone year free of impairment between episodes, and‘¿�persistent'if impairment was more persistent thanthis.

Personality functioning. Personality disorder wasnot reassessed at follow-up because there are nostandard criteria that allowed this diagnosis to bemade in the absence of data on life history, and toinclude information covering the time of the firststudy would have made predictions tautological.Rather, the assessment at follow-up was based ona systematic assessment, using predefined criteria,of impairment in the patient's functioning in work,intimate relationships, day-to-day coping and patternsof social life outside the family, regardless of thepresence or otherwise of episodic psychiatric symptoms. ‘¿�Episodic'impairment was rated if there wasrole impairment in one or more areas for less thanhalf of the assessment period. ‘¿�Single-role'impairment was rated if there was impairment in a singlerole area for more than half of the time, and ‘¿�major'impairment was rated if there was impairment inmore than one role area for one-half or more of theassessment period.

Comparabilityofdiagnosticschemesatbothpoints

Episodic disorders

It was necessary to check the diagnoses in the originalstudy because of the differences in the symptom list

Men Women Total(n=57) (n=74)(n=131)Episodic

disordernoepisodicdisorder21 512psychosis12

1815depression497061anxiety

state16 711otherdisorders2 0 1

y2= 13.16, d.f.=4,P=O.O11Personalitydisorderno

disorder53 7867dramaticgroup37 1826avoidant/dependent10

4 7x2=9.75 d.f.=2,P=0.008Alcoholismnone73

9284heavydrinking9 78alcoholism18

1 8y2= 12.09 d.f.=3, P=0.007

EpisodicdisorderPsychosisDepressionAnxietyNoneMen':

no.728912Personalitydisorder:%none716167-dramatic

group29291192avoidant/dependent-11228Women:

no.135254Personalitydisorder:%none928280-dramatic

group01520100avoidant/dependent8400

318 QUINTON ET AL

and method of diagnosis between those data and thePSE. This was done first by checking whetherCATEGO produced comparable diagnoses on thefollow-up data when restricted to the symptomsincluded in the FIS measure. This proved to be thecase although, naturally, the Index of Definitionfrom the PSE was not comparable. Following this,CATEGOwas applied to the original data and theresulting diagnoses compared with the originalassessments. The differences between the twoschemes were then examined. Discrepancies arosebecause: CATEGOdid not give a diagnosis becauseof insufficient symptoms; the primaryFIS diagnosiswas alcoholism or personality disorder, categoriesnot assigned by cATEclo;FIS diagnoses took accountof psychotic symptoms occurring during the currentepisode but before the three-month assessmentperiod on which symptoms were rated; andassignment to anxiety disorders or depressive illnessvaried because of the weight given to depressed moodin the two schemes.

All patients with discrepant diagnoses werereviewed and diagnoses changed to accord with PSEprinciples. Some patients with personality disordersor alcoholism did not receive episodic diagnoses.

Personality disorders

Recent discussions of the classification of personalitydisorders have suggested that the schizoid andanxious diagnoses should be treated as manifestations of their associated episodic disorders, andthe remaining disorders grouped into two broadcategories (Rutter, 1987; Tyrer, 1988), the firstcategory including actively disruptive and manipulative disturbances in personal relationships(‘dramatic' disorders), and the second coveringdisorders characterised by passive, immature orinadequate social-role performance (an ‘¿�avoidant/dependent' group).

The files of all patients were reviewedand patientsreclassified according to this scheme. Personalitydisorder was rated if there was evidence of persistentimpairment in at least two of three areas: intimate(‘marital') relationships, work, and relationshipswith friends. In nearly all cases there was evidencethat pervasive problems had been present fromadolescence, but personality disorder was also ratedif the patient's adult functioning showed evidenceof marked pervasive problems before the originalstudy, in the absence of episodic illness. Patients withalcoholism were not assigned to personality disordersunless they met the criteria, but all of the 11 patientswith alcoholism also received a designation ofpersonality disorder.

Table 1Revised diagnostic classifications (%)

The distributions of the revised original diagnosesfor the 131 patients in the follow-up study are givenin Table 1. Men were significantly more likely thanwomen to be given diagnoses of personalitydisorder,and for this to be the only diagnosis they received.Women were more likely to be diagnosed as depressedand men as showing anxiety states but the frequencyof psychoses was similar for both sexes.

Results

The overlap between episodic andpersonality disorders

The overlap of episodic and personality disorders atthe initial study key contact is given in Table 2.

Table 2Overlapof episodicandpersonalitydisorders

No statisticallysignificantdifferencesby diagnosis.1. One man with a diagnosis of ‘¿�other'(compulsive gambling) omitted.

EpisodicdiagnosisPsychosis

DepressionAnxietyNonePsychiatric

episodesinpastfive years:no.15 58911none:

%33 675655episodic:%33 17110persistent:

%33 16 33x2=1153 d.f.=6,P=0.0745Role

impairmentinpastfive years:no.15 581012none:

%20 51408episodic:%20 8100single

role:%13 22033persistent:%47 19 50 68

@2=20.93,d.f.=9, P=0.012

31915-20 YEAR FOLLOW-UP OF PSYCHIATRIC PATIENTS

In this table ‘¿�psychosis'includes schizophrenia,puerperal psychoses and bipolar disorders. Unipolaraffective psychoses were assigned to the ‘¿�depressed'group because of the generally mild nature of theirpsychotic features. There were no statisticallysignificant differences between the episodic classifications on whether patients also showed personalitydisorder, but for all episodic diagnoses men weremore likely than women to receive a designation ofpersonality disorder (NS).

Outcome at 15—20years

Deaths

Sixteen of the total original sample of 137 patientshad died by the time of follow-up. Three of thesedeaths occurred during the original study period,including the only established suicide. One furtherpatient died in circumstances that suggested thepossibility of self-harm, but he was within the groupof six patients excluded from the follow-up. Threeother deaths appeared to be due to the physicalconsequences of alcohol abuse. The remainder werenatural deaths for a variety of physical ailments.

Episodic diagnosis

Patients in the depressed group showed a significantly better outcome than the other three groups onthe recurrence of episodic symptoms and on theprobability of pervasive and persistent social impairment (Table 3). The low cell numbers for the otherdiagnostic group make firm conclusions difficult, butit is noteworthy that a third of patients with anxietydisorders did poorly on both measures. Conversely,apart from the psychotic group, half or more of

Table 3Outcome of episodic disorders

patients experienced no impairing episodes of illnessin the five years before the follow-up interview. Thepatients originally diagnosed as having personalitydisorders without accompanying episodic disordershowed significantly poorer social functioning thanpatientswithanyotherdiagnoses,withthemajorityshowing persistent and pervasive role impairment.Strikingly, in nearly half of the cases this pervasiveimpairment was accompanied at follow-up by clearepisodic symptoms.

The consistency of diagnoses across time wasexamined using the CATEGO-derivedfollow-up diagnoses for those patients who had an episode at anIndex of Definition level of 5 or higher in the 12months before interview, regardless of the level ofhandicap, together with summary diagnostic ratingsmade from the interview data or from case notes forepisodes in the five years but outside the past year.The majority of these cases were patients with chronicconditions controlled by medication. Patients onwhom information was only available from childrenor spouses were not included in this analysis. Theresults, by diagnosis, were as follows.

Schizophrenia/paranoid state. Of the eight patientswith schizophrenia, five retained that diagnosis atfollow-up, one was diagnosed as having a neuroticdepression, and two were free of disorder. Theoriginal diagnosis in the subject with depression atfollow-up had been of a paranoid state, as was alsothe case for one of the two in this group who werefree of symptoms. Follow-up data were available ontwo of the three patients with puerperal psychoses,one of whom had no disorder and one of whom wasclassified as suffering an anxiety state.

Bipolar disorder. Three of the four patients in thisgroup received diagnoses at follow-up: two were onlithium and chronically handicapped by their condition, and one other was classified as depressed byCATEGO, but an examination of the case material

indicated that his diagnosis had not changed, a factconfirmed by his continuing use of lithium.

Unipolar depressive psychosis. Only two of theseven patients in this group showed any disorder atfollow-up: one showed an anxiety disorder and theother alcoholism without other symptoms.

Unipolar depression. Twenty out of 48 depressedpatients received follow-up diagnoses. Eleven showedthe same disorder as at first contact, five had anxietydisorders, and one was alcoholic. One developedchronicand profoundlyhandicappingschizophrenia,one had had a recent episode of mania, and one wasclassifiedas having a borderline psychotic state. In bothof these last two cases CATEGOassigned them to newdiagnoses on the basis of brief but ratable incidents ofelevated mood and hallucinatory experiences.

No personalitydisorderPersonality disorderStatistical

significancex2d.f.P%

withpersistentillnessatfollow-upOtherdiagnosispsychosis33%

(4/12)33%(1/3)0.471NSdepression7%(3/45)46%(6/13)9.1710.005anxiety29%(2/7)50%(1/2)0.081NS%

with persistent role impairmentatfollow-upOriginaldiagnosispsychosis42%

(5/12)68%(2/3)0.021NSdepression7%(3/45)62%(8/13)20.2310.0001anxiety43%(3/7)67% (2/3)0.011NS

320 QUINTON ET AL

Anxiety and panic. Four of the nine patients withanxiety diagnoses originally had had episodes in thefive years before the follow-up. Only one of thesewas classified as anxious. The other three were placedin the neurotic depression group.

In summary, although the great majority ofpatients with disorders at follow-up showed illnessessimilar in type to their original diagnoses, there wasonly modest stability over time in the classificationsmade within the neurotic depression and anxietydisorder groups.

Personality disorders

The poor outcome for personality disorders wasconfirmed when all patients with a personalitydisorder, whether accompanied by an episodicdisorder or not, were considered. Only 17010of thosewithout personality problems showed pervasive orpersistent role impairment, whereas over half of thedramatic group and all of those in the small groupof non-dramatic personality disorders did so. Nonein this group, and only 16% in the dramatic group,were free of impairment at follow-up compared withover half of those without a rating of personalitydisorder (x2= 27.50, d.f. = 6, P< 0.0001). Patientswith alcoholism also had poor outcomes, butthe poor outcomes for personality disorder werenot explained by the overlap with alcoholism.Of the eight alcoholics for whom outcome datawere available, three had died from alcoholrelated problems, and of the remaining five,two showed persistent handicap in a single rolearea and three showed pervasive handicap. However,the proportion of those in the ‘¿�dramatic'groupbut without alcoholism who showed pervasivehandicap (50%) was comparable to that of thealcoholics.

Episodic disorders and personality disorders

The next question was the extent to which thepoor outcomes among the episodic diagnoses wereexplained by associated personality disturbance(Table 4).

For the patients in the psychotic or anxiety groupsthe absence of personality disorder did not decreasethe risk of episodic illness 15—20years later. Thefmdings were different for the depressed group. Onlyone-quarter of the patients in this group withoutpersonality disorder had some episodic symptomsover the previous five years and only 7% showedpersistent illness. When personality disorder waspresent, the likelihood of further episodes of illnesswas high (46%) and at a similar level to the othergroups.

Finally, it was necessary to check whether thepoorer prognosis for those with personality disorders was simply a reflection of severity orchronicity of disorder more generally. A check onthe outcome of episodic disorders in the absence ofpersonality disorder showed no significant associations with the length or severity of the originalpresenting episode. Therefore, it seems unlikelythat the poorer outcomes from personality disorderare to be explained simply as consequences of thelength or clinical severity of the associated episodicillness.

Overall outcome of episodic illness

In order to have a complete picture of the prognosisof disorders and the extent of differences betweendiagnoses, it is necessary not only to consider thepersistence of impairment, but also to examinedifferences in the extent of the remission of allsymptoms. To do this, a summary outcome measure

Table4Episodic disorders and outcome with and without personality disorders

disorders) (%)

PsychosisAnergicorAnxiousAnxiety(n= 12)psychoticdepression(n =8)depression(n=

29)(n=26)We))1742012Predominantly

well8195025Intermediate33314525Poor428538

32115-20 YEAR FOLLOW-UP OF PSYCHIATRIC PATIENTS

was constructed for the five years before the followup, including the presence of minor symptoms, theuse of psychotropic drugs, and alcohol misuse (overand above alcoholism), as well as the ratings ofpsychiatric episodes and role impairment. Patientswere rated as: well if there were no problems on anyof these measures; predominantly well if they werewithout role impairment but reported some minorsymptoms; intermediate if there was definite roleimpairment with or without psychiatric symptomsbut for less than half of the past five years, or morepersistent problems in a single role area —¿�usuallymarriage; and poor if there was persistently poorsocial functioning in more than one role area, withor without episodic symptoms, or if death was dueto suicide or to physical illness related to alcoholism.Comparisons on this measure were restricted to thepatients without personality disorder because it wasso strongly predictive of a poor outcome, regardlessof the episodic diagnosis.

It is clear from Table 5 that the conclusions fromthe comparisons on the individual measures ofpsychiatric episodes and role impairment also applywhen attention is focused on the association betweenthe original diagnosis and complete remission ofsymptoms or the occurrence of minor symptomsonly. The depressed group were significantly morelikely to be free of impairment at follow-up thaneither the psychotic group or those with anxietydisorders. Further, when the disorders in thedepressed group were divided into those with andwithout an admixture of anxiety, over two-fifths ofpatients with disorders not involving anxiety werecompletely free of symptoms over the five yearsbefore follow-up. This included the small group ofpatients with umpolar affective psychoses. Theoutcome was different for those with an admixtureof anxious symptoms, none of whom showedcomplete remission. The overall outcome for theseanxious depressed patients was generally similar tothose with unipolar affective psychoses, except that

Table 5Diagnosis and overall outcome (excluding personality

complete remission was more likely for the latter.The patients with anxiety disorders were significantlymore likely than these two groups combined to havea poor outcome and somewhat less likely to showcomplete remission or to remain predominantly well.

Discussion

Three main conclusions are suggested by these data.First, they provide strong evidence for the value ofa categorical diagnosis of personality disorder basedon pervasive abnormalities in social functioning, andfor a broad vision of such disorders into ‘¿�dramatic'and ‘¿�dependent/avoidant'types. Personality disorderproved to be a particularly strong predictor of therecurrence of depression. There are several competingmodels for this association (Klein ci a!, 1993), andfurther research needs to focus on testing the possiblecausal hypotheses. Second, the findings confirm theimportance of taking broader aspects of socialfunctioning into account when assessing the outcomeof all psychiatric disorders (Shepherd ci al, 1989).Third, although there were too few patients in somediagnoses for firm conclusions, the findings supportthose from other studies concerning the outcome forepisodic disorders.

Patients with schizophrenia or paranoid statesshowed a somewhat better outcome than commonlysupposed, with little evidence for a progressivedecline in functioning from the levels evident 15—20years earlier. The four patients with bipolar illnessshowed generally poor functioning. Three-quartersof patients in each of these diagnostic groups weremaintained on medication. The outcome for patientsin the depressed group lay between those reportedby Nystrom (1979) and those in the Lee & Murray(1988) and Kiloh eta! (1988) series. These differencesmay be explicable through sampling criteria as wellas the placement of the bipolar disorders in the‘¿�psychotic'group in our analyses. Unlike the othertwo series, our sample contained both in- and outpatients, omitted the chronically ill and patientswithout children. Thus, it almost certainly containeda smaller proportion of severely ill patients. Thisinterpretation is consistent with the data on suicide.Thus, even if the one possible suicide in our series isincluded and the comparison restricted to depressiveillness, the rate in the full sample (2°lo)was muchlower than those reported by Lee & Murray (1988)and by Kiloh ci al(l988), where the proportions were10°loand 7°lo,respectively.

However, sampling variations do not explain allthe differences. Although the four patients withbipolar affective disorders in our series had pooroutcomes, the patients with unipolar depressive illness@2=26.51,d.f.=9, P=0.O02.

322 QUINTON ET AL

of psychotic intensity or with anergic depression hadthe best outcome of all the episodic illnesses,provided that personality disorder was not alsopresent. These findings do not accord with Lee &Murray's or with Kiloh et al's fmdings. In both thesestudies endogenous depression had a poorer prognosisthan neurotic illness, although in the Maudsley(former) series the prediction was much poorer iftrait neuroticism was present, whereas in the Sydney(latter) series personality features were predictive forneurotic but not endogenous depression. Thesedifferences may reflect variations in the inclusioncriteria for endogenous illness as well as differencesin initial severity between our own and the other twoseries.The differencesaccordingtothepresenceorabsence of adverse personality features are also likelyto be explained by variations in measurement, especiallyof personality functioning. Thus, 26°loof ourdepressed group were rated as having personalitydisorder using criteria based on pervasive socialdysfunction, compared with 27°loin the Maudsleyseries rated on DSM—III(APA, 1980) definitions,and no less than 64% in the Sydney series accordingto ICD—9criteria (WHO, 1978).

There were too few patients with bipolar disordersin our sample to allow any conclusions to be drawnconcerning the psychotic-neurotic continuum. Nevertheless, it should be noted that the outcome for thosewith bipolar illnesses was markedly different fromthose with unipolar depressions of psychotic intensitywho, indeed, were the group with the best prognosis.The difference between these two diagnoses appearedto remain whether or not personality disorder wasor was not present. The absence of personalitydisorder appeared to make no difference to thegenerally poor outcome of the bipolar group,whereas the unipolar disorders had a poor prognosisonly when personality disorder was also involved.

Anxiety and outcome

The prognosis for those with depressive disordersthat did not have an anxious component was morefavourable than for those disorders with an admixture of generalised anxiety. However, the courseof the latter illnesses was also relatively benign.Although none in this group was free of symptomsin the five years before follow-up, they mostlyshowed a recurrence of only minor symptoms, orsome short-lived episodes of role impairment duringthis period. The prognosis for both the anergic andthe anxious depressions was markedly poorer whenthere was associated personality disorder, a fmdingin accord with the findings from the Sydney series.Strikingly, 38% of the small number of patients with

anxiety or panic disorders showed a poor outcomeat follow-up, regardless of the presence or absenceof personality malfunction, and as poor as that forthe psychoses. That this is not a chance fmding issuggested by the similarly poor outcomes in the studyby Krieg ci a! (1987).

The nosological implications of the differences inoutcome according to the presence or absence ofanxiety are unclear. The findings from the existinglong-term follow-up studies do not yet provide theevidence to allow a choice on whether the componentof generalised anxiety is a personality substrate ora comorbid episodic disorder. At present, however,the evidence points more towards the first of thesealternatives. The follow-up of Maudsley depressivepatients, for example, showed few differences inoutcome overall across subtypes of depressive illness,but a poorer outcome for melancholia whenassociated with trait neuroticism (Duggan ci a!,1991). This finding is broadly consistent with ours.Nevertheless, this issue will not be clarified until thereare long-term studies with more sophisticatedmeasurement of state and trait anxiety and theirrelationship to personality functioning.

Course of personality disorders

Finally, we can consider the question of theimprovement in function for those with a designationof personality disorder 15—20years earlier. It isclear that those whose personality functioning wascharacterised by passive, avoidant or inadequate rolefunctioning, had a particularly poor outcome, all thepatients in this group showing pervasively poor socialfunctioning at follow-up. The avoidant and dependent characteristics of this group are similar tothose of the “¿�generalneurotic syndrome―describedin the two-year follow-up of neurotic patients byTyrer ci a! (1992), and also predictive of very pooroutcomes. Compared with this ‘¿�avoidant/dependent'group, the outcome for those with a designation of‘¿�dramatic'personality disorders was more varied.Just over half of them also showed poor functioning,and a small proportion, 14°lo,of this group (fourpatients) were rated ‘¿�well'or ‘¿�predominantlywell'on overall outcome. Improvement did not appear tobe related to whether there was also an episodicdiagnosis initially. This makes it less likely thatthe patients who improved were misclassified onpersonality disorder. Rather, it may be that theimprovement reflects the attenuation of personalitymalfunction with age for patients in the ‘¿�dramatic'group, a process suggested for some disorders of thistype (McGlashan, 1986).However, it may also be thatsocial factors were implicated in the improvements in

32315-20 YEAR FOLLOW-UP OF PSYCHIATRIC PATIENTS

HUBER,G., GROSS,G., SCHLTFTLER,R., et a! (1980) Longitudinalstudies of schizophrenic patients. Schizophrenia Bulletin, 6,593—605.

KENDELL, R. E. (1989) Clinical validity. Psychological Medicine,19, 45—55.

KILOH,L. G., ANDREWS,G. & NEILSON,M. (1988) The long-termoutcome of depressive illness. British Journal of Psychiatry, 153,752—757.

KLEIN, M. H., KUPFER, D. 3. & SHEA, M. Y. (eds) (1993)

Personalityand Depression:A Current View. New York:Guilford Press.

KRIEG, J-C., WITrCHEN, H-U. & VON ZERSSEN, D. (1987) Anxiety

disorders: a long-term prospective and retrospective follow-upof former patients suffering from an anxiety neurosis or phobia.Acta PsychiatricaScandinavica,76, 36—47.

LEE, A. S. & MURRAY, R. M. (1988) The long-term outcome ofMaudsley depressives. British Journal of Psychiatry, 153,741—751.

MCGLASHAN,T. H. (1986) The Chestnut Lodge follow-up studyIII: longterm outcome of borderline personalities. Archives ofGeneral Psychiatry, 43, 20-30.

NYSTROM, 5. (1979) Depressions: factors related to 10 year

prognosis. Acta Psychiatrica Scandinavica, 60, 225-238.RUTTER,M. (1987) Temperament, personality and personality

disorder.BritishJournalof Psychiatry,150, 443-458.& QumrroN,D. (1984)Parentalpsychiatricdisorder:effects

on children. Psychological Medicine, 14, 853—880.SHEPHERD, M., WATF, D., FALLOON, I., et a! (1989) The natural

history of schizophrenia: a five-year follow-up study of outcomeand prediction in a representative sample of schizophrenics.Psychological Medicine (monograph suppl. 15).

SToNE, M. H. (1993) Long-term outcome of personality disorders.BritishJournalof Psychiatry,162, 299—313.

Tyaen, P. (1988) What's wrong with DSM-III personality disorders?Journal of Personality Disorders, 2, 281-291.

SIEVEWRIGHT,N., FERGUSON,B., et a! (1992) The generalneurotic syndrome: a coaxial diagnosis of anxiety, depressionand personality disorder. Acta Psychiatrica Scandinavica, 85,201-206.

VIZE, C. & TYRER, P. (1994) The relationship between personality

and other psychiatric disorders. Current Opinion in Psychiatry,7, 123—128.

WEISSMAN,M. M., PRUSOFF, B. A. & Kutaxw@a, G. L. (1978)Personality and the prediction of long-term outcomein depression. American Journal of Psychiatry, 135,797—800.

WING, .1. K. & HAlLEY, A. (1972) EvaluatIng a CommunityPsychiatric Service: The Camberwell Psychiatric Register1964—1971.London: Oxford University Press.

COOPER, 3. E. & SARTORIUS, N. (1974) The Description and

Classificationof PsychiatricSymptoms:An Instruction Manualfor thePSEand CategoSystem.London: CambridgeUniversityPress.

WORLDHEALThORGANIZATION(1978) The Ninth Revision of theInternational Classification of Diseases and Related HealthProblems (ICD-9). Geneva: WHO.

Ziw.atv.wi, M. & CORYELL, W. (1989) DSM-III personalitydisorder diagnoses in a nonpatient sample. Archives of GeneralPsychiatry, 46, 682-689.

these patients and in the reduction of pervasivemalfunction to not more than a single-role problemarea for nearly half of those with dramaticpersonality disorders. Social influences on the courseof both episodic and personality disorders areconsidered in a companion paper.

Acknowledgements

Early stages of this study were supported by the Foundation forChild Development and the W. T. Grant Foundation, and thefollow-up by the Medical Research Council. We are most gratefulto all the study members for sharing their experienceswith us andto Caroline Amina, Carol Bell, Mary McNutt, Christina Shearerand Carol Stimson for interviewing in the follow-up study.

References

AKISICAL, H. S., HIRSCHFELD, R. M. A. & YERVANIAN, B. 1. (1983)The relationship of personality to affective disorders. Archivesof General Psychiatry, 40, 801—810.

AMERICAN PSYCHIATRIC ASSOCIATION (1980) Diagnostic andStatisticalManual of Mental Disorders(3rd edn, revised)(DSM-II1-R). Washington, DC: APA.

ANDREWS,G., NEILSON,M., Hutrr, C., ci a! (1990) Diagnosis,personality and the long-term outcome of depressives. BritishJournal of Psychiatry, 157, 13—18.

BIEHI, H., MAURER, K., SCHUBART, C., et a! (1986) Prediction ofoutcome and utilization of medical services in a prospective studyof first onset schizophrenias: results of a five year follow upstudy. European Archives of Psychiatry and NeurologicalSciences,236, 139—147.

CHARNEY, D. S., NELSON, C. 3. & QUINLAN, D. M. (1981)

Personality tests and disorder in depression. American Journalof Psychiatry,138, 1601-1604.

COSTA, P. T. & MCRAE, R. R. (1986) Personality stability and itsimplications for clinical psychology. Psychology Review, 6,407—423.

DOCHERTY, 3. P., FIESTER, S. 3. & SHEA, T. (1986) Syndrome

diagnosis and personality disorder. In American PsychiatricAssociationAnnual Review, Vol. 5 (edsA. J. Frances& R. E.Hales), pp. 315—355.Washington, DC: American PsychiatricPress.

DEAKE, R. E. & VAILLANT, G. E. (1988) Longitudinal view ofpersonality disorder. Journal of Personality Disorders, 2,44-48.

DUGOAN, C. F., Len, A. S. & Muen,@v, R. M. (1990) Does

personality predict long-term outcome in depression? BritishJournal of Psychiatry, 157, 19-24.

& (1991) Do different subtypes ofhospitalised depressives have different long-term outcomes?Archives of GeneralPsychIatry, 48, 308—312.

FRIEDMAN, C. .1, SHEAR, M. K. & Fw4cas, A. (1987) DSM-IIIpersonality disorders in panic patients. Journal of PersonalityDisorders,1, 132—135.

GRenR, H. 3. & CAWLEY,R. H. (1966) Natural History of Neuroticillness. Mervyn Archdall Medical Monograph 3. Sydney:Australian Medical Association.

David Quinton, PhD, Department of Social Work, School for Policy Studies, University of Bristol;Lesley Gulliver, BA, Michael Rutter, FRS, MRC Child Psychiatry Unit, Institute of Psychiatry, London

Correspondence: Dr David Quinton, Department of Social Work, School for Policy Studies, University of Bristol, 8Woodland Rd, Bristol BS8 1TN. Fax: 0171 708 5800

(First received9 June 1994,final revision2 November 1994, accepted 10 November 1994)