measuring disordered personality functioning: to love and to work reprised
TRANSCRIPT
Measuring disordered personalityfunctioning: to love and to work reprised
Introduction
�Perhaps the word diagnosis can be eventually dispensed
with in favor simply of assessment of functioning� (1).
Individuals satisfying diagnostic criteria for onepersonality disorder (PD) commonly satisfy cri-teria for several other PDs (2, 3). This can reflecttrue comorbidity or inability of clinicians andclassification systems to distinguish accuratelybetween distinct PDs. Current diagnostic descrip-tors for PD assess both disordered function (DF)and personality style, with descriptors of DFgenerally broad and non-specific, risking furtherconfounding. Thus, if several PDs include aparticular aspect of functioning, such as impulsiv-ity, then its presence will build to comorbidity evenif the associated-personality �styles� are truly dis-tinct.
Difficulties in diagnosing and measuring the PDsvalidly have encouraged an argument that person-ality style and DF should be measured independ-ently (4, 5). This would clarify whether most or allPD personality styles are associated non-specific-ally with DF, or whether each personality styleshapes specific disordered functioning.Another argument for measuring DF as an
independent construct is to assist definition of PDstatus. DSM-IV and ICD-10 criteria for personal-ity disorders emphasize personality traits. How-ever, such traits may be intrinsically maladaptive,or only maladaptive above a certain imprecisethreshold or in specific contexts.
Aims of the study
To define PD primarily on the basis of DF, andthen assess personality style as a secondary
Parker G, Hadzi-Pavlovic D, Both L, Kumar S, Wilhelm K, Olley A.Measuring disordered personality functioning: to love and to workreprised.Acta Psychiatr Scand 2004: 110: 230–239.�BlackwellMunksgaard 2004.
Objective: Current limitations to diagnosing and measuring thepersonality disorders encouraged a set of studies seeking to provide analternate approach to modeling and measuring disordered personalityfunction.Method: A large set of self-reported descriptors of disorderedpersonality function were factor analyzed in a sample of patients withclinician-diagnosed personality dysfunction, generating 11 lower-orderand two higher-order constructs. Subjects and non-clinical controlsalso completed a measure of personality styles underpinningformalized personality disorder groupings. Properties of the refinedself-report (SR) measure were assessed in an independent sample ofpatients with a clinically diagnosed personality disorder.Results: Limitations in �cooperativeness� and �coping� formed thehigher-order constructs defining disordered personality function, withthese constructs relevant to all personality styles. Analyses of SR,corroborative witness (CW) and clinician-rated data in an independentsample supported measuring disordered personality function by ourderived 20-item SR measure, and exposed limitations to clinician-basedassessment.Conclusion: Study findings build to a multi-axial strategy formeasuring personality disorder, involving separate dimensionalassessment of both disordered personality function and of personalitystyle.
G. Parker1, D. Hadzi-Pavlovic1,L. Both2, S. Kumar2, K. Wilhelm1,A. Olley11School of Psychiatry, University of New South Wales,Mood Disorders Unit, Black Dog Institute, and 2Prince ofWales Hospital, Sydney, Australia
Key words: personality disorder; personality;disordered functioning
Prof G. Parker, Euroa Unit, Prince of Wales Hospital,Randwick 2031, Sydney, Australia.E-mail: [email protected]
Accepted for publication February 6, 2004
Acta Psychiatr Scand 2004: 110: 230–239Printed in UK. All rights reserved
Copyright � Blackwell Munksgaard 2004
ACTA PSYCHIATRICASCANDINAVICA
230
element. To pursue this approach (and as outlinedin Fig. 1), we refine putative disordered personalityfunction constructs to develop a model of DF,examine whether DF is linked specifically or non-specifically to differing PD personality styles,compare data from PD subjects and normalcontrols, and test the validity of our refinedmeasure. Finally, we compare the capacity ofpersonality style and DF to distinguish betweensubjects with and without a PD.
Material and methods
Measures
Disordered function (DF). From our detailed litera-ture review (6) we identified 17 constructs histor-ically defining DF such as �instability under stress�,�impulsivity� and �inflexibility�. A total of 141descriptors were written for the constructs [in aself-report (SR) format], and reviewed by experi-enced clinicians in terms of content, relevance andambiguity. We administered the descriptors to twonon-clinical samples, totaling 146 subjects, toreduce the item set to a more practical number.A provisional factor analysis identified �non-coop-erativeness� and �disagreeableness� constructs assynonymous, �maladaptability� and �lack of humor�had only one loading item each, while the construct�failure to form and maintain relationships� did notgenerate a discrete factor. Items for the last three
constructs were deleted together with items quan-tified as synonymous with other items (correlationsof more than 0.70). A reduced set of 82 itemsgenerated our current study’s DF measure. Eachitem allowed four responses, �definitely false�,�mostly false�, �mostly true� and �definitely true�,scored dimensionally (0, 1, 2, or 3 respectively).
Personality style (STYLE). A previous study (7)developed descriptors of 15 formalized PDs, inclu-ding all those listed in DSM-IV and in ICD-10(paranoid, schizoid, schizotypal, narcissistic, anti-social, borderline, histrionic, avoidant, dependent,obsessive and depressive), some previously listed indiagnostic manuals (passive-aggressive, self-defeat-ing, sadistic) and an �anxious� PD category. Ana-lyses of large SR and observer-rated databasesgenerated the 142-item STYLE questionnairemeasuring personality style scores for each of the15 putative PD categories. Each item allowed fourresponses, �definitely false�, �mostly false�, �mostlytrue� and �definitely true�, scored dimensionally(0, 1, 2, or 3 respectively).
Global judgment of disordered personality function(GLOBAL JUDGEMENT). This measure reques-ted clinicians to �rate the extent to which theindividual’s personality style and function are suchthat they and/or others around them suffer as aconsequence�. Options allowed �no�, �slight�, �mod-erate� and �distinct� limitations to function.
Fig. 1. Flow chart of component studies.
Measuring disordered personality functioning
231
Functional domains (DOMAIN). This measurerequired clinicians to rate the subject�s functionacross five �domains� (work, and intimate, peer,family and work relationships), with four-pointscales providing options of �functional�, and �poss-ibly�, �probably� or �definitely� dysfunctional.Summed scores generated a total DOMAIN score.
Clinician-rated personality disorder style (PD VIGN-ETTE). This measure required clinicians to ratethe extent to which the patient�s personality stylecorresponded to succinct descriptors of each of the15 PD categories (i.e. �not at all�, �to some degree�,�reasonably well� and �completely�).
Samples
Clinical sample. We approached some 50 psychia-trists to recruit suitable patients. Eligibility criteriawere: (i) clinician-judged presence of either a PDor significant personality traits affecting function;(ii) age 18–65 years; (iii) not currently having apsychotic disorder, mania, severe depression, sig-nificant cognitive impairment or a distinct drug oralcohol problem; and (iv) competence in English.The psychiatrists completed GLOBAL JUDG-MENT, DOMAIN and PD VIGNETTE measuresfor eligible subjects and invited them to completean anonymous SR questionnaire (for their returnby pre-paid mail) collecting demographic informa-tion as well as DF and STYLE measure data.We recruited 249 subjects referred by 33 psychi-
atrists. The study strategy, respecting patientanonymity and choice in returning questionnaires,disallowed formal determination of the responserate.
Control sample. Research staff were asked to selectfriends and close colleagues judged over time asnot showing any evidence of a PD. Invitees wererequested to complete the STYLE and DF meas-ures for return by pre-paid mail. To protectanonymity we did not record any identifyingdetails, and did not determine the response rate.This sample consisted of 67 subjects.
Measure validation sample. This next study wasdesigned and initiated after completing data ana-lyses of the clinical and control samples. Hospitalpatients meeting the same criteria as for the clinicalsample were requested to: (i) complete the DFmeasure; (ii) have a relative or friend complete acorresponding observer-rated DF measure �bestdescribing�…how �X…usually or generally feels orbehaves, as they have been over the years, and notjust recently� (returned directly to the research
team); and (iii) be interviewed by two clinicalresearch staff.The two raters, a psychiatrist (SK) and a
psychologist (AO), administered a semistructuredinterview based on the Adult Personality Func-tioning Assessment measure (8), assessing lifetimefunctioning across six domains: work, intimaterelationships, relationships with family and friends,work relationships, non-intimate social contacts,and coping and negotiating styles when stressed.Using six-point dimensional scales with structuredanchor points, each rater independently assessedthe extent to which the patient evidenced �cooper-ativeness� and �effectiveness� across each domain,with summed scores generating total interviewer-rated �non-cooperativeness� and �non-coping� scoresrespectively. Thirty-five subjects (and 26 corrobor-ative witnesses) were recruited.
Results
Refining the constructs to develop a model of disorderedfunctioning
Clinical sample members had a mean age of39.1 years (SD 11.6 years), a mean of 15.0 years(SD 3.6 years) of education, and a slight femalepreponderance (52%). Only 27% were married,18% divorced or separated and 0.4% widowed.Just over one-half were in full-time or part-timework, 13% were unemployed, 18% pensioners, 7%undertaking home duties and 6% students.STYLE scores for respondents were converted to
percentage scores for each putative PD category toallow standardized prevalence estimates. A score of100% for a PD would represent all respondentsscoring 3 (i.e. �definitely true�) for every item relevantto that PD. Percentage scores were: depressive(54%), anxious (53%), obsessive (48%), avoidant(41%), borderline (40%), schizotypal (38%), self-defeating (37%), schizoid (37%), paranoid (36%),dependent (35%), histrionic (34%), negativistic(29%), antisocial (28%), narcissistic (26%), andsadistic (20%). Males scored higher (P < 0.05) onthe narcissistic scale and females scored higher onthe borderline, anxious and depressive scales. Therewere no sex differences for the remaining STYLEscores.We compared relevant SR STYLE and psychiat-
rist-rated PD VIGNETTE scores. Pearson correla-tion coefficients were: borderline (0.46), depressive(0.36), paranoid (0.35), avoidant (0.32), anxious(0.31), schizoid (0.28), passive-aggressive (0.27),narcissistic (0.27), dependent (0.24), histrionic(0.23), sadistic (0.20), antisocial (0.20), self-defeat-ing (0.19), schizotypal (0.18) and obsessive (0.07). A
Parker et al.
232
mean coefficient of 0.26 across the 15 PD categoriessuggested only slight overall agreement betweensubjects and psychiatrists in rating PD styles.Initial factor analyses of the 82 DF items
identified 15 factors, although not all of thesewere well supported (with some factors having lessthan four items, and some items loading <0.03).Deleting these factors and items resulted in a set of67 items and 11 factors. Factor analysis of scalescores generated from the first-order factorsidentified two second-order factors. These solu-tions provided the basis for a confirmatory factoranalysis (CFA) in this sample – an a priori patternfor 11 first-order and two second-order factors.The CFA was carried out in LISREL and thesolution gave a moderately good fit to the data.Based on Hoyle (9), the RMSEA statistic (0.74)was slightly better than a marginal fit; the GFIstatistic (0.65) was poor; while the CFI statistic(0.96) was very good. Given that we had a largenumber of items in a relatively small sample, weinterpreted reasonable support for the proposedfactor model.Table 1 reports factor loadings and associations
between item scores for each of the lower-order DFconstructs and clinician-rated DOMAIN andGLOBAL JUDGMENT scores, with associationsranging from non-existent to slight, suggesting thatseverity of lower-order DF constructs did notcorrelate with clinical judgments of PD severity.As for the STYLE measure, DF percentage
scores were calculated for the 11 constructs toallow standardized prevalence estimates of: �insta-bility under stress� (58%), �self-defeating� (50%),�pessimism� (49%), �ineffectiveness� (45%), �lowself-direction� (44%), �impulsivity� (42%), �inflexib-ility� (42%), �disagreeableness� (35%), �not learning
from experience� (33%), �non-empathic� (29%) and�uncaring to others� (27%). The only sex difference(P < 0.05) was for males to score lower on the�non-empathic� scale.Based on the regressions of the second-order
factors on the items from the CFA we took the 10items with the highest weights for non-coopera-tiveness and summed item scores to form anon-cooperativeness scale. We similarly generateda 10-item non-coping scale. Contributing items(listed in the Appendix) came from seven of the 11first-order factors. The two scales correlatedr ¼ 0.97 and 0.98 respectively with the factorscores for the second-order factors, and with eachother (r ¼ 0.56 and 0.64 for scales and factorsrespectively). There was no sex difference for non-coping (males ¼ 11.6, females ¼ 11.2, t ¼ 0.54,d.f. ¼ 307, P ¼ 0.59) but there was for non-coop-erativeness (males ¼ 7.9, females ¼ 6.0, t ¼ 3.66,d.f. ¼ 307, P < 0.001).There was no difference in returning positive
GLOBAL scores for either non-coping (F ¼ 2.27;d.f. ¼ 2, 97; P ¼ 0.11) or non-cooperativeness(F ¼ 0.41; d.f. ¼ 2, 97; P ¼ 0.67). The correlationsbetween DOMAIN and the two scales were quitelow (r ¼ 0.18 and r ¼ 0.16). These two resultsindicate negligible agreement between self-reportedand clinician-rated levels of disordered functioning.We next examined if DF (as measured by scales
for lower-order and higher-order DF constructs)was associated specifically or non-specifically withdiffering PD styles as measured by the 15 STYLEscores. Table 2 shows relevant correlation coeffi-cients. Lower-order DF scale scores were associ-ated with STYLE scores, supporting a modelwhereby dysfunction was non-specifically associ-ated with the separate PD styles. The non-specific
Table 1. Factor loadings in confirmatory second-order factor analysis of DF constructs andcorrelation of lower-order factor scores withclinician-rated DOMAIN and GLOBAL JUDGMENTscores
Lower-order DF construct
Loadings on higher-order factors
Correlation between DF Scaleand clinician-rated measure of
PD severity
Factor 1(non-coping)
Factor 2(non-cooperativeness) DOMAIN
GLOBALJUDGMENT
Disagreeableness – 0.96 0.32** 0.18Inflexibility 0.73 – 0.14 0.12Uncaring to others – 0.88 0.20* 0.08Non-empathic 0.90 0.20* 0.09Ineffectiveness 0.94 – 0.21* 0.19*Self-defeating 0.96 – 0.27** 0.27**Failure to learn fromexperience
0.88 – 0.04 0.02
Impulsivity 0.62 – 0.28** 0.06Pessimism 0.91 – 0.01 0.07Instability under stress 0.77 – 0.00 0.08Lacking self-direction 0.99 – 0.16 0.12
Pearson correlation significant at *0.05 level and **0.01.
Measuring disordered personality functioning
233
associations might reflect either the interdepend-ence of the DF constructs or our STYLEdescriptors truly being non-specifically linked toDF. Scores for the two higher-order DF con-structs also linked non-specifically with STYLEscores (Table 2), with associations almost invari-ably stronger with �non-coping� than with �non-cooperativeness� scores.We then contrasted data from samples with and
without a personality disorder. On all STYLEmeasures and lower-order DF scales, patientsreturned significantly higher scores ( P < 0.001)than the non-clinical controls. On the 10-item DFscales, the patients returned significantly higher�non-coping� (13.1 vs. 5.1, t ¼ 10.9, d.f. ¼ 314,P < 0.001) and �non-cooperativeness� (7.6 vs. 4.2,t ¼ 5.8, d.f. ¼ 314, P < 0.001) scores. Receiveroperating characteristic (ROC) analyses showed�non-coping� to have an area under curve (AUC) of0.87 and an optimal cutoff [minimizing thesensitivity (Se) and specificity (Sp) difference] ofnine or more, at which point Se ¼ 0.78 andSp ¼ 0.81. For �non-cooperativeness� the AUCwas 0.73 at the optimal cutoff of 6 or more(Se ¼ 0.67 and Sp ¼ 0.66). Some 60% of patientswere positive on both cutoffs, compared with 15%of controls.We next examined a central question: in
modeling the PDs, should descriptors of person-ality style and/or of DF be prioritized? Thisquestion was approached by first determining therelative capacities of the 15 personality STYLEscores and the two 10-item DF scale scores todistinguish patients and controls, using discrimi-nant analysis.In one analysis, only the two higher-order DF
scales were entered. The standardized weights (0.95and 0.10) indicated that scores on the �non-coping�scale discriminated more than �non-cooperative-ness� scale scores. The sensitivity and specificity inpredicting true �cases� and true �non-cases� wasrelatively high (76 and 82% respectively, canonicalR ¼ 0.60).In a second analysis, the set of 15 STYLE
scores only was entered using a stepwise proce-dure, with only two STYLE scores being retainedin the final equation (anxious and depressive, withweights of 0.48 and 0.64 respectively). The sensi-tivity and specificity in predicting true �cases� andtrue �non-cases� was again high (81 and 91%respectively, canonical R ¼ 0.60). The prominenceof anxious and depressive STYLE scores raisedthe possibility that groups might have beendistinguished more by mood disturbance than byPD status. We therefore repeated the analyses,after removing �anxious� and �depressive� scores Ta
ble2.
Corre
lations
ofself-ratedPD
STYLEscores
with
self-reportDF
scores
andwith
clinician-ratedjudgments
STYLE
Lower-order
DFscores
Clinicianjudgments
Disagreeableness
Inflexibility
Uncaringto
others
Non-
empathic
Ineffectiveness
Self-
defeating
Failure
tolearnfro
mexperience
Impulsivity
Pessimism
Instability
under
stress
Lacking
self-direction
10-item
non-coping
10-item
non-cooperativeness
Clinician-rated
DOMAINscore
Clinician-rated
PDVIGN
ETTE
Paranoid
0.69
0.63
0.43
0.41
0.58
0.70
0.47
0.53
0.57
0.56
0.63
0.67
0.59
0.39
0.39
Schizoid
0.54
0.49
0.55
0.41
0.56
0.67
0.51
0.33
0.43
0.63
0.63
0.66
0.50
0.40
0.21
Schizotypal
0.60
0.50
0.46
0.39
0.55
0.69
0.41
0.49
0.48
0.53
0.58
0.63
0.52
0.36
0.14
Narcissistic
0.59
0.50
0.35
0.35
0.37
0.46
0.30
0.53
0.32
0.26
0.40
0.41
0.51
0.24
0.30
Anti-social
0.69
0.60
0.43
0.41
0.61
0.67
0.51
0.64
0.57
0.51
0.64
0.68
0.59
0.43
0.24
Borderline
0.60
0.55
0.36
0.28
0.67
0.72
0.57
0.60
0.63
0.61
0.70
0.73
0.48
0.35
0.44
Histrionic
0.57
0.45
0.30
0.28
0.48
0.54
0.41
0.61
0.47
0.37
0.48
0.52
0.44
0.26
0.24
Avoidant
0.48
0.46
0.38
0.29
0.63
0.66
0.46
0.34
0.56
0.61
0.58
0.66
0.40
0.20
0.32
Dependent
0.41
0.45
0.24
0.21
0.66
0.61
0.56
0.46
0.62
0.51
0.61
0.67
0.33
0.17
0.26
Obsessive/
compulsive
0.45
0.59
0.35
0.29
0.49
0.58
0.43
0.29
0.52
0.50
0.47
0.53
0.38
0.15
0.06
Anxious
0.49
0.55
0.32
0.23
0.69
0.66
0.58
0.40
0.81
0.62
0.62
0.71
0.38
0.17
0.40
Depressive
0.47
0.49
0.41
0.26
0.73
0.76
0.62
0.42
0.62
0.80
0.71
0.78
0.40
0.20
0.36
Sadistic
0.59
0.51
0.34
0.35
0.34
0.43
0.28
0.54
0.33
0.29
0.40
0.39
0.54
0.23
0.21
Self-defeating
0.56
0.51
0.40
0.29
0.72
0.77
0.58
0.48
0.60
0.71
0.70
0.77
0.46
0.35
0.21
Negativistic
0.63
0.59
0.37
0.37
0.63
0.72
0.53
0.57
0.56
0.57
0.70
0.72
0.55
0.28
0.22
Parker et al.
234
from the predictor set of STYLE variables. Threeof the 13 STYLE scores were retained: borderline,sadistic, and self-defeating (standardized weightsof 0.52, )0.26 and 0.68 respectively) and withreduced sensitivity and specificity (77 and 87%,canonical R ¼ 0.56).When the DF scores for �non-coping� and �non-
cooperativeness� were entered first and then the 13STYLE scores were entered stepwise, only�self-defeating� was added to the equation (withstandardized weights of 0.47, 0.06 and 0.59respectively). The resulting sensitivity (92%) andspecificity (65%) gave no advantage.As an additional test of capacity to distinguish
samples, a series of ROC curves was calculated forthe STYLE scale and two higher-order DF scales,and for composite scales where the best compositewas calculated as described by Su and Liu (10).Table 3 reports sensitivity, specificity and the non-parametric AUC, indicating how well each predic-tor performed as a diagnostic test. Analyses of thehigher-order DF scales alone showed that the �non-coping� scale was superior to that of �non-cooper-ativeness� scale, and that combining the scales didnot produce any advantage over use of the�non-coping� scale alone. An analysis of individualSTYLE scores showed that all were highly effectivein differentiating patients from controls (in partic-ular, borderline, anxious, depressive and self-defeating styles). The performance of the combinedSTYLE scores was slightly superior to the
combined DF scales (with respective AUCs of0.94 and 0.89).We next sought to validate the DF measure in
the independent sample of 35 subjects (meanage ¼ 35.1 years; 63% female) with personalitydysfunction. Of these, 26 had a nominated familymember (for example, parent, spouse, de factopartner) or close friend complete an observer-ratedDF questionnaire. The mean �non-coping�DF scalescores of the sample was higher than that generatedin the development sample by the patients (16.4 vs.13.1), while �non-cooperativeness� scores werecomparable (8.4 vs. 7.6).For validation, we first examined agreement
between patient’s SR and CW rated scores on theDF scales. �Non-coping� scale scores were similar(SR ¼ 17.0; CW ¼ 15.9, t ¼ 0.95, d.f. ¼ 24, ns)and moderately correlated (r ¼ 0.61, P < 0.001),as were �non-cooperativeness� scale scores(SR ¼ 8.8; CW ¼ 8.8, t ¼ 0.08, d.f. ¼ 24, ns;r ¼ 0.41, P ¼ .04). Secondly, correlation betweeninterviewers in rating lifetime �non-coping� and�non-cooperativeness� was high (Pearson correla-tions of 0.70 and 0.77 respectively, P < 0.001).However, when we examined the extent to whichthose interviewer-based ratings correlated withSR-rated and CW-rated DF scores, agreementwas poor. The correlation of mean interviewer-judged rating of �non-coping� with the relevant SRscale was 0.34 (ns) and 0.36 (ns) with the CWscale. The correlation of the mean interviewer-judged rating of �non-cooperativeness� with therelevant SR was 0.33 (ns) and 0.25 (ns) with therelevant CW scale.In summary, we found stronger agreement for
the SR measure with CW reports than withinterviewer ratings. This supports the validity ofthe SR strategy.
Discussion
Disordered functioning – conceptually broad, empiricallyparsimonious
We commenced assuming that disordered person-ality function is likely to be expressed in multipleways. A detailed literature review (6) identified 17constructs, and we expected that subsequentrefinement would still result in a sizeablenumber. However, results suggested a moreparsimonious model of DF. For example, whenthe constructs were reduced to 11, each wasmoderately to highly associated with each other.This led to a secondary analysis defining justtwo higher-order constructs – �non-coping� and�non-cooperativeness�, with scores on the two 10-
Table 3. Capacity of two disordered function (DF) scale scores and 11 style scoresto differentiate clinical and control samples, using receiver operating characteristic(ROC) analyses
Analysis Test Se Sp AUC AUC 95%CI
I DF scales – individualNon-coping 0.83 0.77 0.89 0.82–0.94Non-cooperativeness 0.67 0.66 0.74 0.69–0.79
II DF scales – combined 0.80 0.80 0.89 0.82–0.93III PS scales – individual
Paranoid 0.80 0.79 0.85 0.78–0.89Schizoid 0.77 0.75 0.85 0.79–0.90Schizotypal 0.76 0.82 0.84 0.78–0.89Narcissistic 0.72 0.66 0.738 0.68–0.77Antisocial 0.77 0.80 0.84 0.78–0.88Borderline 0.85 0.84 0.89 0.82–0.94Histrionic 0.69 0.71 0.77 0.71–0.81Avoidant 0.82 0.77 0.87 0.80–0.91Dependent 0.76 0.80 0.850 0.79–0.90Obsessive/compulsive 0.67 0.77 0.80 0.74–0.85Anxious 0.83 0.79 0.90 0.83–0.94Depressive 0.86 0.82 0.91 0.84–0.95Sadistic 0.65 0.71 0.72 0.67–0.77Self-defeating 0.82 0.80 0.91 0.83–0.95Negativistic 0.71 0.71 0.82 0.76–0.86
IV PS scales – combined 0.86 0.86 0.94 0.85–0.98
Se, sensitivity; Sp, specificity; AUC, area under curve for ROC.
Measuring disordered personality functioning
235
item derived scales being moderately associated(r ¼ 0.45). Interdependence of the two constructsmay reflect methodological issues, with associa-tions between them being dependent on othervariables (e.g. severity) or response biases (e.g.social desirability).Alternatively, disordered personality function
might be expressed – and therefore measured –very simply as failure to cooperate and failure tocope. There is historical support for such a model– with Freud (11) defining full maturity asrequiring only two markers: �Lieben und arbei-ten�� (to love and to work, p. 238). �To love� wasassumed to refer to the capacity to make andmaintain intimate relationships, and assumescooperation. �To work� referred not merely tobeing employable but involved in �work-produc-tiveness�.Our strategy for refining the constructs of
disordered personality function relied on factoranalysis, a strategy that has been criticized asimposing a structure rather than discovering one(12), and its limitations need to be recognizedhere. In fitting a CFA we examined how well aprevious finding �carried across� differing samples.We considered that a two-factor higher-ordersolution provided both the �best fit� of the dataas well as the potential for greater flexibility infuture applied research. However, the need toretain �non-cooperativeness� can be challenged, inthat �non-coping� was superior to �non-coopera-tiveness� in every relevant analysis. In thediscriminant analyses, it was not only superiorto �non-cooperativeness� but made its contributionredundant.However, until the model has been tested
further, we suggest however that �non-coopera-tiveness� be retained. Discarding it now couldnarrow definition of personality dysfunction to acoping domain and ignore a (cooperative) con-struct central to many definitions of PD. Forexample, Livesley et al. (13) defined PD as atripartite failure of (i) the self-system, with failureto establish stable and integrated representation ofself and others; (ii) adaptive function in interper-sonal relationships; and (iii) prosocial and cooper-ative relationships, while Rutter (14) held that �apervasive persistent abnormality in maintainingsocial relationships� underpinned many categor-ized PDs.
Models of PD
How well does our two-construct model corres-pond to other models of PD? We have alreadynoted a previous emphasis on �cooperativeness�.
Several authors (15, 16) have also considereddisordered personality function to be manifestedaccording to a coping paradigm. Millon (17)argued for three principal features differentiatingnormal and pathological behaviour (functionalinflexibility, self-defeating circles and tenuous sta-bility under stress) and later an additional set offive parameters (causing personal discomfort,reduction of opportunities, ineffective function,non-adjustment to the environment, and causingdiscomfort to others). Our analyses included allsuch constructs and, while demonstrating theirutility as markers, defined them as lower-orderconstructs contributing principally to the higher-order construct of �non-coping�. Livesley et al. (13)emphasized inflexibility and inadequate perform-ance in �the universal life tasks of identity, attach-ment, intimacy and affiliation�, also noting thattheir definition did not differ greatly �from Freud’sdefinition of mental health as the ability to workand to love�.Despite starting with a very wide set of indic-
ative constructs, our final two-construct model isstrikingly similar to that emerging from Clonin-ger’s studies in which character scales of low�self-directedness� and low �cooperativeness� wereproposed as the core features of PD (18). OurCFA empirically established the dominating con-tribution of the higher-order constructs, with, forexample, scores on the lower-order construct�lacking self-direction� loading 0.99 on thehigher-order �non-coping� construct, and thelower-order construct �disagreeableness� loading0.96 on the higher-order �non-cooperativeness�construct.
Specificity of disordered functioning
If such a parsimonious model of DF is valid,scores on the DF measure should be associatedwith all, rather than merely some, PD categories.We did find some instances of specificity: forexample, narcissistic and obsessive/compulsivepersonality STYLE scores were somewhat lessstrongly associated with �non-coping� and �non-cooperativeness� scores than other STYLE scores.Their less clear relationship with DF is supportedby the observation (19) that narcissistic andobsessive PDs are the �two disorders that oftenallow patients to succeed in the Western cultures�(p. 268). However, the overall pattern described inTable 2 supports a non-specificity model. Thiscould reflect the nature of disordered personalityfunction, in that all PDs are evidenced by limita-tions in coping and cooperativeness. It is alsopossible that the non-specific pattern may reflect a
Parker et al.
236
higher-order factor, such as severity of disorderedpersonality.
Validity of self-rating
Although our patient sample was derived frompsychiatrists recruiting patients judged clinically ashaving distinct limitations in personality function,we found little agreement between SR and clini-cian-rated data. The correlation between psychia-trists� judgments of personality styles (PDVIGNETTE) and patient-generated STYLEscores was only 0.26, and patient-generated DFand style scores correlated poorly with psychiatrist-rated DOMAIN scores. Despite high inter-rateragreement between two professionals in our valid-ation study assessing �non-coping� and �non-coop-erativeness�, we found weak agreement betweeninterviewers� scores and both subjects� SR and CWscores. However, SR scores were correlated withCW scores on the DF measure, supporting the SRstrategy. Such results suggest that external raters,whether trained interviewers or treating psychia-trists, have limited capacity to assess DF – hardlysurprising when external raters are observingindividuals in a socially constrained contextrather than in vivo.
Diagnostic classification
In considering how study results link with theformal diagnosis and classification of PDs, weshould note key historical issues, with Livesley (5)describing three classificatory phases. First is thepre-DSM-III phase, dominated by clinical descrip-tions within phenomenological and psychoanalyticfields. Empirical research in the 1970s led to thesecond (DSM-III) phase, with PDs placed on aseparate axis and having diagnostic criteria,encouraging further empirical analyses. Livesleyargued that the field is entering a third, post-DSM-III/IV phase, reflecting limitations to the DSMmodel. Although the alternative ICD-10 classifica-tion differs in some details, its PD conceptualiza-tion is similar to DSM-IV.The DSM-IV model is categorical and includes
10 formalized PDs organized into three higher-order clusters, a categorical classification consis-tent with traditional medical approaches seeking toidentify specific diseases and syndromes. However,the multiple problems presented by a categoricalsystem are well recognized (20). The necessity tojudge disorders as either �present� or �absent� resultsin arbitrary distinctions between �normal� and�abnormal� personality. In clinical application,diagnoses of DSM PD categories have low reliab-
ility and validity. Structured and semistructuredinterviews have been developed because of suchlimitations to clinical assessment, but no scale hasemerged as the �gold standard�. It has long beenunclear whether subject or informant report is themore valid, with choices influenced by the subject’sinsight, how well the informant knows the subject,the nature of the trait being measured and thepurpose of the assessment (e.g. clinical assessment,research).We earlier noted that current PD definition
reflects an amalgam of descriptors of personalitystyle and of DF. By examining the capacity ofseparate measures of these components to discrim-inate subjects judged a priori as possessing or notpossessing significant personality problems, wesought to determine which construct was thesuperior discriminator. After stripping the �depres-sive� and �anxious� PD STYLE predictors frominitial analyses, to overcome distinctions beingmore driven by mood state, both STYLE and DFparameters had almost equivalent capacity todiscriminate between patients and controls. How-ever, the STYLE descriptors were developed fromrelevant DSM and ICD descriptor sets, so manywere �contaminated� by descriptors of personalityfunction. If we had alternatively stripped STYLEdescriptors of their �function� components, ourstudy would have been so much at variance withthe existing DSM model as to make comparativejudgments impossible. Nevertheless, as PD dys-function was predicted strongly by both style andfunctioning components, the use of both parame-ters is supported.The results encourage a research model testing
the utility of a two-tiered model of PD which firstemphasizes the likelihood of disordered personalityfunction (e.g. �definite�, �probable�, �possible�,�absent�), and secondarily provides descriptors ofpersonality style. This is consistent with a proposalby Livesley (5) that PD be diagnosed by DF alone,and noted within axis I of the DSM system, whilepersonality styles would be positioned on axis II.In contrast to the current categorical diagnosticmodels, both tiers would ideally be dimensionallybased, consistent with the view (21) – and empiricalfindings (13) – that PDs are maladaptive orextreme expressions of common personality traitsrather than being qualitatively different fromnormal personality function.The secondary component in the proposed two-
tier model could emphasize either �normal� or�abnormal� personality styles. As an exampleof the former, the five-factor model of person-ality (22) proposes dimensions of neuroticism,extroversion, openness to experience, agreeableness
Measuring disordered personality functioning
237
and conscientiousness. Emphasizing �normal� per-sonality dimensions would allow a bridge betweenthe two disparate dimensional and categorical�worlds�, and between psychological and psychi-atric research. The alternative second-tier optionwould be to derive prototypic phenotypes ofdistinct personality styles as observed by clinicians.Prototypes could respect those captured in currentand past DSM and ICD systems or other categor-ical models (19), with an individual’s adherence tothe prototype also being dimensionally measured.A third �mixed model� is another option, and mightassist model refinement. The classificatory accu-racy and clinical utility of the two-tiered modelcould then be compared with more traditional PDassessment strategies.
Conclusions
Our review and the studies reported in this paperadvance a new model for measuring and research-ing the PDs, and which might have classificatoryutility. The model is dimensional and two-tiered,separating descriptors of DF from descriptors ofpersonality style. The model also allows fordisjunctions between tiers, in that some individualsmay have extreme personality �styles� but functionwell, and vice versa. For example, current descrip-tors of antisocial personality disorder emphasizethe characteristics of the �failed sociopath�, but inreality descriptors that include personality style aswell as level of function can be associated witheither �success� or �failure�.The two-tiered model overcomes limitations of
current models attempting to capture separatedomains of style and function within a singlefield, and has the potential to provide moremeaningful information to clinicians and toresearch endeavors.
Acknowledgement
This study was funded by an NHMRC Program Grant(222308) and an Infrastructure Grant from the NSW Depart-ment of Health. The authors thank the many psychiatrists whoassisted with study recruitment.
References
1. Millon T, Davis R. Disorders of Personality. DSM-IV andBeyond. New York: John Wiley and Sons Inc., 1996.
2. Pfohl B, Coryell W, Zimmermann M et al. DSM-IIIpersonality disorders: diagnostic overlap and internal
consistency of individual DSM-III criteria. Compr Psy-chiatry 1986;27:21–34.
3. Joffe RJ, Regan JJ. Personality and depression. J PsycholRes 1988;4:279–286.
4. Parker G. The etiology of personality disorders: a reviewand consideration of research models. J Pers Disord 1997;11:345–369.
5. Livesley WJ. Introduction. J Pers Disord 2000;14:1–2.6. Parker G, Both L, Olley A. et al. Defining disordered
personality functioning. J Pers Disord 2002;16:503–522.
7. Parker G, Hadzi-Pavlovic D. A question of style: refiningthe dimensions of personality disorder style. J Pers Disord2001;15:300–318.
8. Hill J, Harrington R, Fudge H. et al. Adult PersonalityFunctioning Assessment (APFA): an investigator-basedstandardised interview. Br J Psychiatry 1989;155:24–35.
9. Hoyle RH. Confirmatory factor analysis. In: Tinsley HEA,Brown SD, eds. Handbook of Applied Multivariate Sta-tistics and Mathematical Modelling. San Diego: AcademicPress, 2000; 465–497.
10. Su JQ, Liu JS. Linear combinations of multiple diagnosticmarkers. J Am Stat Assoc 1993;88:1350–1355.
11. Erikson EH. Childhood and Society, 2nd edn. London:Random House, 1995.
12. Frances A, Widiger TA. Methodological issues in per-sonality disorder diagnosis. In: Milloon T, Klerman GL,eds. Contemporary Directions in Psychopathology:Toward the DSM-IV. New York: Guilford Press, 1986:381–400.
13. Livesley WJ, Jang KL, Vernon PA. Phenotypic and geneticstructure of traits delineating personality disorder. ArchGen Psychiatry 1998;55:941–948.
14. Rutter M. Temperament, personality and personalitydisorder. Br J Psychiatry 1987;150:443–458.
15. Kruedelbach N, McCormick RA, Schulz SC et al. Impul-sivity, coping styles, and triggers for craving in substanceabusers with borderline personality disorders. J PersDisord 1993;7:214–222.
16. Vollrath M, Alnaes R, Torgersen S. (1998). Coping stylespredict change in personality disorders. J Pers Disord1998;12, 198–209.
17. Millon TA. Theoretical derivation of pathological per-sonalities. In: Millon T, Klerman GL, eds. ContemporaryDirections in Psychopathology: Toward the DSM-IV. NewYork: Guilford Press, 1986:639–669.
18. Svrakic DA, Whitehead C, Przybeck TR et al. Differentialdiagnosis of personality disorders by the seven-factormodel of temperament and character. Arch Gen Psychiatry1993;50:991–999.
19. Westen D, Shedler J. A prototype matching approach todiagnosing personality disorders: toward DSM-V. J PersDisord 2000;14:109–126.
20. Parker G, Barrett E. Personality and personality disor-der: current issues and directions. Psychol Med 2000;30:1–9.
21. Widiger TA, Costa PT. Personality and personality disor-ders. J Abnorm Psychol 1994;103:78–91.
22. Costa PT, McCrae RR. Revised NEO Personality Inven-tory (NEO PI-R) and NEO Five-Factor Inventory(NEO-FFI) Professional Manual. Florida: PsychologicalAssessment Resources, Florida, 1992.
Parker et al.
238
AppendixTen items loading on each of the higher-order scales, and the lower-order scales from which each was derived.
Lower order scale
Non-cooperativenessEven when I have to, I am unable to get along with family or people at work AgreeablenessI am generally described as a nice person* AgreeablenessI can be somewhat difficult in dealing with others AgreeablenessIn general, I will listen to and understand the other person's point of view* AgreeablenessFriends see me as cooperative and agreeable* AgreeablenessPeople at work see me as cooperative and agreeable* AgreeablenessI tend to be very understanding of other people's feelings and problems* EmpathyI am generally ready and willing to lend an ear* EmpathyPeople see me as good-hearted* Caring for othersPeople who know me well would describe me as a caring person* Caring for others
Non-copingI seem to fail more often than I succeed in life Self-defeatingMy personality often causes me to lose out Self-defeatingI know I cope poorly with things EffectiveWhen things go wrong I am generally able to bounce back* EffectiveI feel confident in my ability to size up and deal with any situation* EffectiveI learn from the mistakes I make* Learns from experienceI am really resourceful in tackling problems* Self-directedOthers see me as a reliable person* Self-directedI feel I have little control over where my life is headed Self-directedI feel like I am going around in circles in life Self-directed
*Reverse scoring.
Measuring disordered personality functioning
239