Multivariate statistical methods and problems of classification in psychiatry

A commentary on Effective Connectivity Between the Orbitofrontal Cortex and the Precuneus Differentiates Major Psychiatric Disorders: Results from a Transdiagnostic Spectral DCM Study

Translational psychiatry has been a hot topic in Neurosciences research. The authors present a commentary on the relevant findings from a transdiagnostic study applicable to clinic practice. Additional discussion on conceptual and clinical insight into this current broad line of research is explored in the integration of multi-level paradigm in Psychiatry research.

Dimensional assessment of personality pathology in patients with eating disorders

This study examined patients with eating disorders on personality pathology using a dimensional method. Female subjects who met DSM-IV diagnostic criteria for eating disorder (n = 136) were evaluated and compared to an age-controlled general population sample (n = 68). We assessed 18 features of personality disorder with the Dimensional Assessment of Personality Pathology - Basic Questionnaire (DAPP-BQ). Factor analysis and cluster analysis were used to derive three clusters of patients. A five-factor solution was obtained with limited intercorrelation between factors. Cluster analysis produced three clusters with the following characteristics: Cluster 1 members (constituting 49.3% of the sample and labelled 'rigid') had higher mean scores on factors denoting compulsivity and interpersonal difficulties; Cluster 2 (18.4% of the sample) showed highest scores in factors denoting psychopathy, neuroticism and impulsive features, and appeared to constitute a borderline psychopathology group; Cluster 3 (32.4% of the sample) was characterized by few differences in personality pathology in comparison to the normal population sample. Cluster membership was associated with DSM-IV diagnosis -- a large proportion of patients with anorexia nervosa were members of Cluster 1. An empirical classification of eating-disordered patients derived from dimensional assessment of personality pathology identified three groups with clinical relevance.

The assessment of personality in young adulthood: Data on a normative sample

The Personality Assessment Schedule has been modified for use with young adults living in the community. It has been administered to a population sample of 129 subjects aged 24-26 years. It proved possible to achieve a reliable assessment of personality trait. Further, utilising both clinical and statistical approaches personality types were defined. The dimension reflecting types of personality also proved reliable. Guidelines are provided as to what constitutes deviant scores on the elicited personality dimensions.

Prediction of response to electroconvulsive therapy. Preliminary validation of a sign-based typology of depression

In an open study of 35 depressed patients, psychomotor disturbance, assessed by the CORE rating system, significantly predicted response to ECT in patients with a 'retarded' but not those with an 'agitated' type of illness. In multivariate analyses, psychomotor disturbance was the sole predictor of response. This preliminary study suggests that the CORE system, which has been shown to differentiate endogenous and non-endogenous depressive disorders, has predictive validity in relation to ECT response.

The factor structure of manic rating scales

Separate factor analyses were performed on interview and ward behaviour ratings of 81 manically disturbed patients. The principal components analyses produced a dimensions of manic disturbance separating elation from aggressiveness. For the interview scale the rotated factors were (a) thought disturbance, (b) overactive and aggressive behaviour and (c) elevated mood and vegetative symptoms. For the ward behaviour scale the rotated factors were (a) motor and speech disturbance, (b) aggressiveness and (c) unrealistic expansiveness of mood.

Endogenous depression as a construct: a quantitative analysis of the literature and a study of clinician judgements

We review the concept of endogenous depression historically and undertake quantitative analyses of representative factor analytic studies. We also report an empirical study, isolating symptoms and signs associated with a clinical diagnosis of psychotic/endogenous depression, made by a large number of clinical raters assessing 300 depressives. The quantitative analyses and the study of practising psychiatrists agreed in delineating depressive type and together suggest a more restricted construct of endogenous and psychotic depression than has been held historically. Key clinical features isolated were severity, retardation, delusions and paranoid features, non-reactivity and non-variability in mood, while vegetative features appeared to lack relevance.

Alcoholism--assessment and diagnosis

Ninety-two patients with alcoholism were interviewed on various aspects of alcohol use. Factor analysis of the data revealed five factors: withdrawal symptoms; help seeking behaviour; pathological drinking; legal difficulties; family problems. The analysis suggested the possibility of heterogenity in this clinical sampling with alcoholism.

Can categorical and dimensional views of psychiatric illness be distinguished?

Bimodality in a distribution of symptoms is often claimed to be convincing evidence that a disorder is categorical, a discrete disease entity, rather than the extreme on a continuous dimension. However, using concepts from contemporary psychometric theory it is shown that bimodality can arise from the dimensional viewpoint. In fact, contrary to the usual belief, bimodality would be expected to occur in many research contexts if the dimensional alternative were correct.

The numerical approach to classification: a medical application to develop a typology for facial pain

Applications of numerical taxonomy in medicine are relatively rare. We describe an application in which a new typology is developed for the classification of pain distribution in patients with temporomandibular pain dysfunction syndrome. The methodology of the numerical approach is presented and key issues are highlighted, especially relating to validation of taxonomic structure. From the dendrogram obtained by Ward's hierarchical method, five pain groups are distinguished which relate systematically to anatomy and are clinically meaningful. Cluster structure is subjected to a battery of reproducibility tests and is found to be broadly reproducible for different algorithms and sub-samples of subjects. Finally, the new typology is shown to have heuristic value in providing fresh insight into the nature of a complex syndrome.

Prediction of outcome in depressive illness by the Newcastle Diagnosis Scale. Its relationship with the unipolar/bipolar and DSM-III systems

The Newcastle scores of a group of 64 and subsample of 52 severely depressed inpatients were not normally distributed. Evidence for discontinuity in these distributions was adduced from the contrast in outcome between the endogenous and neurotic patients thus defined, the endogenous consistently doing better than the neurotic group. The unipolar/bipolar system failed to predict different results for endogenous and neurotic patients unless unipolar was subdivided into endogenous and neurotic subgroups. The DSM-III criteria for major depression and melancholia failed to identify subgroups of differing prognoses.

Subtypes of panic attacks and ICD-9 classification

No single ICD-9 category corresponds to panic disorder (DSM-III). To investigate whether patients with panic attacks can be identified by means of ICD-9, 97 patients with three panic attacks within 3 weeks were recruited from various medical centers, and were classified independently according to DSM-III and ICD-9. The ICD-9 diagnoses were scattered over a broad range of categories, and it was impossible to identify patients with panic disorder in this manner. Anxiety state, affective psychosis, and depressive neurosis were the most frequent ICD-9 diagnoses. The boundary between affective psychosis on the one hand and anxiety state and depressive neurosis on the other hand was validated by present and previous symptomatology and by cluster analysis. The boundary between anxiety state and depressive neurosis could not be validated in this way. Correspondingly, modifications of the ICD-9 classifications are proposed.

The validity of two Newcastle Diagnostic Scales in the affective disorders. Bimodality and other correlates

Current evidence on the issue of bimodality in the affective disorders is reviewed. A series of 180 consecutive hospital patients with affective disorders were scored on the 1965 and 1972 Newcastle Diagnostic Scales. The 1965 (Carney) scale produced a bimodal distribution which was accounted for by the bipolar and unipolar groups, and some implications are discussed.

An empirical taxonomy of symptom types among divorcing persons

This study generated and assessed an empirical taxonomy of symptom profiles or types in a nonclinical population. Three hundred and thirty men and women who recently had separated from their marital partners were interviewed and completed a variety of instruments. Cluster analysis of factors derived from one such instrument, a symptom checklist, yielded a provisional classification of respondents into nine types. The latter were examined for differences in socio-economic characteristics as well as several variables suggested in the literature to predict or help explain symptom formation: Feelings of control, self-concept, social stressors, and current morale. Results indicated three types to be relatively well off despite the potentially stressful conditions inherent in divorce. The Overwhelmed and the Somatic depressives, in contrast, demonstrated multiple signs of dysfunction. The findings indicate the value of applying empirical classification procedures to at-risk populations in the community.

The relationship between anxiety and depression. A review of some principal component analytic studies

Forty studies published between 1934 and 1977 which used principal component analysis of symptoms, personality or illness features in patients with affective disorders were examined. Factor clusters or dimensions indicative of anxiety and depression were evident in each study. Visual plots on the first 2 components in each of 6 studies selected as being representative of the 40 are included in this paper. The implications arising from semantic confusion are discussed.

Classification and diagnosis of depression in school phobia

Fifty-one school phobic children, aged nine to fourteen years, were assessed for psychiatric diagnosis; this revealed the presence of two clinically meaningful sub-groups--depressed and residual school phobic. A wide range of symptoms were studied to identify those which might prove useful in diagnosing adult-type depression in childhood and early adolescence, both in terms of frequency of symptoms in the depressed group and the extent of the distinction between the two groups. Eleven such key symptoms were identified and based on these, a formula for diagnosing adult-type depression was evolved. The validity of several different ways of classifying the above cases were explored; these covered: kind of disorder; type of onset; adolescence versus pre-adolescence; and sex of the child. However, on only one dichotomy--depression versus absence of depression--were there many significant discriminants. Affective symptomatology of more recent onset was contrasted with pre-morbid personality traits, usually associated with school phobia. There was no evidence to support the concept of 'masked depression' in childhood.

Multivariate analysis of angiographic, histologic, and electrocardiographic data in patients with coronary heart disease

In 61 consecutive patients undergoing aortocoronary bypass grafting, angiographic and electrocardiographic (ECG) changes were studied. Histologic delineation of myocardium was obtained by analysis of transmural biopsy specimens acquired at the time of surgery. The use of principal-component analysis revealed three definite groups of patients. Group I comprised patients with histologic findings associated with severe left anterior descending coronary artery (LAD) stenosis, without abnormal wall motion or ejection fraction. ECG abnormalities were limited to ST changes. Group II comprised patients with severe myocardial cell degeneration with only modest fibrosis associated with severe LAD stenosis and severely impaired wall motion. The incidence of infarction on the ECG was low. Group III patients had important myocardial cell degeneration with severe fibrosis associated with severe LAD stenosis, severely depressed wall motion, and significantly impaired ejection fraction. In this group there was a high incidence of infarction apparent on the ECG. Postoperative follow-up (24 months) showed a total survival of 94.4% in group I, 92.8% in group II, and only 72.7% in group III. This identification of subtypes of coronary artery disease seems to be helpful in estimating patient prognosis after coronary surgery.

Dimensions of psychopathology among alcoholic patients, college students, prison inmates and general psychiatric patients

The Differential Personality Inventory was administered to four samples of subjects: 404 alcoholic patients, 292 college students, 720 prison inmates, and 181 general psychiatric patients. Exploratory factor analysis on the self-report data yielded seven factors in each sample. The six factors, interpreted as common across samples, were: hypochondriacal complaining, blunted affect versus emotional reactivity, interpersonal conflict, depressed withdrawal, cognitive disorganization, and dysfunctional coping. Implications for the empirical development of a multiaxial diagnostic system and the construction of a case-identifying instrument in psychiatric epidemiological research were discussed.

Clinical characteristics of akathisia. A systematic investigation of acute psychiatric inpatient admissions

Assessment of drug-induced movement disorders was carried out regularly on 104 psychiatric patients requiring antipsychotic medication on admission to hospital. The data relevant to motor restlessness were subjected to a principal components' analysis. According to their component scores, patients were then classified into two main groups: an akathisia group and an illness-related-movement group, the former group showing the clinical and pharmacological characteristics expected of akathisia. Clinical features which distinguished between the two groups, and between grades of akathisia severity, were identified, so an objective, phenomenological description of the akathisia syndrome was possible. Our observations suggested two distinct types of acute akathisia; one related to severe parkinsonism and one not. The implications of these findings are discussed.

Classification and diagnosis of depression

There is no general consensus how to classify and diagnose depressive disorders. The main reason is disagreement regarding etiology. This is reflected in the several theoretical models of depression which has been proposed. On the other hand, a reliable way of characterizing groups of patients that can be used by both clinicians and researchers, is necessary both for communication and progress of research. Facing this situation, the inclusion of etiological theories in classification of depression obviously would be an obstacle to the use of such a classification system by clinicians of various theoretical orientation at current. Thus, a classification system which describes the clinical features of the mental disorders comprehensively and at the lowest order of inference necessary to identify the disorder in a reliable way is needed. The third edition of the Diagnostic and Statistical Manual of Mental Disorder (DSM-III) represents such a classification system. By providing operational diagnostic criteria for each diagnosis, DSM-III represents a major achievement of classification in clinical psychiatry and research. This paper will present the classification of depression and the problem of differential diagnosis as outlined in DSM-III. Further, the validity of the DSM-III classification of depression will be discussed with reference to some recent research findings on biopsychosocial and treatment aspects of depressive disorders.

[Principle factors and symptom complexes in delirium tremens, factor and a group analytic study]

Before treatment 80 unselected patients suffering from delirium tremens were examined with regard to 13 psychopathological criteria. For the data a matrix of correlation was computed and it was factor analyzed according to the principal-component method. In consideration of the course of the value-curve two factors were interpreted. The result, however, is an undetailed classification. In addition to that, the data were cluster-analyzed according Ward. The results of the multivariate statistical analysis admit the assumption of two great, though heterogenous groups of symptoms (hallucination/vigilance). Factor I comprises the symptoms, disorder of orientation and consciousness, sweating, agitation and tremor on its positive pole, the duration of the delirant state on its negative pole. Factor II combines paranoid-hallucinatory symptoms, fearful affects and suggestibility on its positive pole, while on its negative, there are happy affect and grand-mal seizures. The bipolarity of this factor and additional diagnoses show that paranoid-hallucinatory symptoms without disorder of consciousness and grand-mal seizures mutually exclude each other. From this a differential therapy treating patients suffering from paranoid-hallucinatory symptoms with neuroleptics (e.g. Haloperidol) can be deduced, while the danger of grand-mal seizures has to be considered when disorders of consciousness appear.

Two types of cross-gender identity

A revision of the typology of male cross-gender identity was carried out by means of formalized, easily replicable methods. The results suggest (1) that there are two discrete types of cross-gender identity, one heterosexual, the other homosexual; (2) that transvestism, and closely related conditions of cross-gender identity, occur exclusively or almost exclusively in heterosexuals; (3) that of the two types of transsexualism distinguished in this study, type A is, in heterosexuals, very rare or completely nonexistent; (4) that (in the course of time) transvestites or borderline transsexuals (defined below) may develop sustained cross-gender identity, as observed by Stoller (1971); (5) that although, according to Hoenig and Kenna (1974), transsexualism by itself is not an anomalous erotic preference, it is (virtually) always either preceded by transvestism or accompanied by homosexuality or cross-gender fetishism.

Have multivariate statistics contributed to classification?

In the last twenty years many workers have applied multivariate statistical techniques to problems of classification in psychiatry. To what extent have the techniques resolved the problems? This personal view will attempt to appraise the current situation.

Identification of the endogenous depressive syndrome based on the symptoms and the characteristics of the course

Using a sample of 198 depressed patients (145 female, 53 male) retrospective histories of the illnesses were collected during a depression-free interval, based on a catalogue containing 38 symptom items and the course of the depression (including the interval personality). A cluster analysis on persons and items filtered out an endogenous depressive item profile, corresponding with the clinical syndrome of patients diagnosed as endogenous depressives in the clinics, although determined without reference to the clinical diagnoses. Our study supports many results from earlier multivariate statistical studies. We consider our data to be an essential contribution towards the establishment of a multiaxial clinical picture.

Bimodality and the nature of depression

Arguments concerning the nature of depressive disorders have involved as a central issue the question of the bimodality or otherwise of the distribution of some variable expressing variation in symptomatology. The implications of a particular type of frequency distribution along this dimension, whether uni- or bi-modal, have been misunderstood by a number of workers, and an attempt is made to clarify the situation.

A symptom profile analysis of antipsychotic drug treatment: nonparametric multidimensional technique

Thirty-six schizoaffective patients were treated with haloperidol plus lithium or haloperidol plus placebo. Standard statistical analysis showed greater improvement in those treated with lithium plus haloperidol, as published previously. The Brief Psychiatric Rating Scale (BPRS) scores were reanalyzed in the present study using a nonparametric technique for multidimensional scaling of symptom profiles. The differential outcome after haloperidol with or without added lithium is determined both by symptom dimension and by type of drug. Added lithium prevents a depressive shift after haloperidol treatment on the dimension of "cognitive disorganization--depressive preoccupation."

The classification of patients into diagnostic groups using cluster analysis

The use of cluster analysis for the classification of patients into diagnostic groups remains controversial. Further evidence is provided for assessing the use of cluster analysis in comparison with other multivariate methods. It is suggested that, when appropriately used, cluster analysis is a convenient tool for developing empirically based diagnostic groupings and that the frequently stated limitations of the technique should not invalidate the obtained groupings in such cases.

Classification of organic brain syndromes by cluster analysis

Cluster analysis was carried out on a sample of 92 patients with behavior disorders caused by degenerative, vascular, (alcohol) toxic, and other diseases of the brain. Rating variables of the AMDP system concerning mental state, social behavior, need for special care, sleep disorders, autonomic, physical, and neurologic symptoms were used in the absence of severe degrees of disordered consciousness such as stupor, coma, delirium tremens, and gross cerebral lesions. Results suggested the existence of four major groups of global cognitive impairment combined with neurasthenia and irritability in the first, hypochondriasis and depression in the second, withdrawal symptoms in the third, and severe disorientation in the fourth. At the seven-group level the groups were further distinguished according to severe withdrawal, amnestic syndrome, and dementia by various social and illness behaviors, sleep-wakefulness pattern, hypo- or hyperactivity, additional physical, and neurologic symptoms. Other minor types of organic brain syndromes were identified as individual cases by hallucinations or other circumscribed cognitive, psychomotor, affect, motivation, personality, and/or behavior disorder, symptomatic manic, or schizophreniform psychosis. The findings lent support to old classifications and new ones of organic mental syndromes (DSM-III).

The WHO Depression Scale. Relationship to the Newcastle Scales

The WHO (World Health Organization) Depression Scale is a new rating scale developed to evaluate the complete clinical history and description of depression. Items of the WHO scale were transformed to cover the items of the two Newcastle scales that are designed to differentiate between endogenous and non-endogenous depressions. In 98 depressed inpatients examined prior to antidepressive therapy only moderate agreement between scores transformed to the two Newcastle scales was found. Also the distribution patterns of the scores on the two scales were different. In patients classified as endogenous on one or both scales a significant relationship between plasma levels of imipramine or clomipramine and antidepressive effect was found. In patients classified on both scales as doubtful - or non-endogenous depressions - no such correlation could be identified. The initial severity of depression, as defined by the Hamilton Depression Scale, was the same in endogenous, doubtful and non-endogenous depression according to the scores transformed to the two Newcastle scales.

Sub-syndromes of tardive dyskinesia

A reliable method for recording the site and duration of purposeless movements was devised. With this method 267 subjects were studied, 182 of whom had been exposed to neuroleptics. The results were submitted to a principal components analysis and 3 movement dimensions emerged. One group of movements resembled a parkinsonian syndrome. The other 2 groups of movements both conformed to the generally accepted criteria for tardive dyskinesia. These groups were: (1) head and neck movements and (2) trunk and limb movements. The possibility of the second and third groups representing clinically relevant sub-syndromes of tardive dyskinesia is discussed.

The structure of depression found in the general population

Principal component analysis was applied to data relating to depressive experiences collected from a general population sample. The structure of depression found in the general population showed strong similarities to that found in the psychiatric population. There were, however, important differences. Four substantial syndromes of depression were identified: 'anxiety depression', 'cognitive depression', 'vegetative depression' and depression characterized by the classic endogenous triad of diurnal variations, early wakening, and weight loss. These syndromes were essentially independent for this sample. Examination of component score distributions suggested that people simultaneously experience differing degrees of each syndrome. It was concluded that depression was most parsimoniously classified in dimensional rather than categorical terms within this sample.

Hyperkinesis comes of age: what do we know and where should we go?

The first heuristic description of the hyperkinetic impulse disorder made its appearance in the literature just over twenty years ago. Although there are many who are made restless and irritable by continuing ambiguities and controversies surrounding the hyperkinetic syndrome, the past two decades have brought considerable maturation in our knowledge of this concept. The present paper sketches the state of the art in regard to hyperkinesis, and draws conclusions about directions for both research and clinical work.