Characterization of Psychotic Conditions: Use of the Domains of Psychopathology Model

Background  Previous factor analytic studies of patients with schizophrenia have consistently demonstrated the presence of 3 psychopathological domains labeled positive, negative, and disorganized. This study examined whether similar domains can be seen in disorders other than schizophrenia, and the degree to which such domains are independent of diagnostic categorization.

Methods  Data from the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (DSM-IV) field trial involving 221 patients with schizophrenia and 189 patients with nonschizophrenia diagnoses were factor analyzed to study the nature of psychopathological domains in the 2 groups. Differential associations between each domain and selected clinical variables were assessed.

Results  Factor analysis yielded a similar 3-factor model of positive, negative, and disorganization domains for patients with schizophrenia as well as other diagnoses. Differential associations found between individual domains and clinical variables (premorbid functioning and negative domain; absence of remissions and disorganization domain) were similar in both schizophrenia and nonschizophrenia groups.

Conclusions  The 3 psychopathological domains previously described in schizophrenia are not specific for that diagnosis. Differential associations found between individual domains and clinical variables were not limited by diagnostic categorization. The results suggest that these domains are not unique to schizophrenia and may each correspond to a discrete pathophysiologic condition.

IT IS GENERALLY accepted that patients grouped under the category of schizophrenia vary widely in their clinical presentation, course, and response to treatment. Much effort has been made to subtype schizophrenia into groups that are clinically more homogeneous. However, relatively little attention has been paid to psychotic conditions grouped under categories other than schizophrenia. The degree to which such categories are distinct from schizophrenia has not been adequately assessed. The DSM-IV uses a categorical-hierarchical model for subtyping schizophrenia.¹ An alternative approach to subtyping schizophrenia, which has received recent attention, is the domains of psychopathology model. In this article, we report on the applicability and validity of this model for psychotic patients classified under diagnoses other than schizophrenia.

In recent years, there has been growing interest in examining the possibility that specific groups of symptoms (domains of psychopathology) within the category of schizophrenia may arise from independent disorders. It has been postulated that co-occurrence of such independent domains in different combinations in different patients may explain the observed clinical heterogeneity of schizophrenia. At the empirical level, identification of such psychopathological domains can be attempted through factor analysis, which groups symptoms on the basis of the correlations among their relative severities. Changes in the severity of an underlying disorder would be reflected in corresponding changes in the severity of symptoms produced by it, while symptoms produced by other pathologic processes would show no correlation. Under this model, a group of symptoms that show a consistent correlation in their degrees of severity can be assumed to arise from the same underlying disorder.

Studies that factor analyzed ratings of patients with schizophrenia by means of the Scale for Assessment of Negative Symptoms (SANS)² and the Scale for Assessment of Positive Symptoms (SAPS)³ have consistently demonstrated the existence of symptom groups corresponding to 3 independent underlying factors.^4-8 These 3 symptom groups have been generally referred to as positive, negative, and disorganization domains. It has been shown that symptoms in each domain may be differentially associated with specific biological, neuropsychological, course, and treatment response variables.^9-16 These data, together with the consistent replicability of the 3 domains in different schizophrenia populations, support the hypothesis that each domain arises from an independent pathologic process.

Given the significant symptom overlap among different psychiatric diagnoses, we believed it would be valuable to study whether these 3 psychopathological domains demonstrated in schizophrenia can also be seen in other diagnoses. Optimally, such a study would use the SAPS and SANS to rate patients, as these are the instruments most commonly used in previous factor analytic studies of patients with schizophrenia. Few studies have made such a comparative analysis. Minas et al¹⁷ studied a heterogeneous group of psychotic patients by means of SAPS and SANS and found domains comparable with those described in schizophrenic patients alone. More recently, Maziade et al¹⁸ compared patients with familial schizophrenia and bipolar disorder and found a similar pattern of 3 domains in both groups. The sample of heterogeneous psychotic patients studied by Minas et al included a substantial number of patients with schizophrenia and schizophreniform disorders, while the study by Maziade et al was restricted to patients with a highly familial pattern. In addition, the sample sizes of nonschizophrenic patients in both studies may be considered small for factor analysis (73 and 87), making interpretation of results difficult.

In this study, we analyzed data obtained for the DSM-IV Field Trial for Schizophrenia and Related Psychotic Disorders,¹⁹ to answer the following questions: (1) Will the 3 psychopathological domains identified in schizophrenia also be seen in patients grouped under other diagnostic categories? (2) Do the individual domains show a differential association with relevant clinical variables? And if they do, are such associations independent of diagnostic groupings? Since the study questions were not formulated at the time of data collection, the raters were unaware of the aims of the study.

The sample consisted of 412 patients from the DSM-IV Field Trial for Schizophrenia and Related Psychotic Disorders. Patients were recruited from 8 sites: 7 sites in the United States and 1 in Mexico. At each site, a recruitment coordinator reviewed consecutive admissions to the inpatient and/or outpatient service participating in the field trial. All patients gave informed consent to participate in the study. At least 1 of the following positive or negative symptoms had to be currently present to at least a mild degree for a subject to be included in the study: delusions, hallucinations, disorganized speech, flat affect, poverty of speech, and passive, apathetic social withdrawal. Information was obtained from subject interviews, current and past medical records, and informant interviews.

All patients were assessed with the field trial instrument, which was composed of several scales, including the SAPS and the SANS. More than 90% of the patients in the sample were also in either interrater reliability or test-retest reliability studies. The reliability for diagnoses and symptom ratings ranged from good to excellent. Although diagnoses were made by several criteria, only diagnoses from the Diagnostic and Statistical Manual of Mental Disorders, Third Edition, Revised (DSM-IIIR) are reported herein.²⁰ The sample included 221 patients with a DSM-IIIR diagnosis of schizophrenia, 189 patients with DSM-IIIR diagnoses other than schizophrenia, and 2 patients with missing diagnoses. Table 1 presents descriptive data on the sample.

The global subscale scores from SAPS and SANS were used for analysis in the present study. Each item was scored on a scale of 0 to 5, with a score of 0 indicating absence of the symptom, 1 indicating questionable presence, and scores from 2 to 5 indicating progressively increasing levels of severity. There were a total of 10 global items: affective flattening or blunting, alogia, avolition/apathy, asociality/anhedonia, catatonia, delusions, hallucinations, disorganized speech (positive formal thought disorder), disorganized behavior, and inappropriate affect. Of these, catatonia was excluded because of a low base rate of prevalence (8%) relative to the other variables. The remaining 9 items were retained for analysis.

Information on some clinical variables other than those used for factor analysis was also available. The following variables were studied to see whether they showed an association with individual domains, and whether such associations were independent of diagnostic groupings: age at first psychiatric symptom; history of remissions; number of hospitalizations; frequency of hospitalizations (duration of illness/number of hospitalizations); fraction of total illness duration spent in the hospital (duration of hospitalization/duration of illness); current, past year, and past 5 years Global Assessment Scale scores; presence of tardive dyskinesia; and premorbid level of functioning. The majority of these variables relate to severity of illness and course and thus were of interest as validators of the domain model. In addition, we were interested in examining the relationship between tardive dyskinesia and the 3 domains, as previous reports indicate an association with the negative domain.²¹

Premorbid functioning was rated on 7 subscales with scores ranging from 0 (good function) to 2 (poor function). For this analysis, scores on the 7 subscales were summed to form 1 composite item with a score range of 0 to 14. In the field trial, history of remissions was scored as either present or absent. For testing its association with individual domains, we excluded patients whose total duration of illness was 2 years or less. This was done to include only those patients whose illness was of sufficiently long duration for a remission to have occurred.

STATISTICAL ANALYSIS

Factor analysis was carried out using the SPSS statistical package (SPSS version 6.1 for Windows, SPSS Inc, Chicago, Ill). The correlation matrix of the variables entered into the analysis is given in Table 2. Principal-component analysis was first conducted, and the scree plot and eigenvalues were used to determine the number of factors to be extracted. An orthogonal rotation (varimax) was conducted to facilitate interpretation of the factors. Separate analyses were carried out for patients given a diagnosis of schizophrenia (n=221) and for those with all other diagnoses (n=189). In addition, subgroup analyses were done for patients with primary mood disorders (n=65) and schizoaffective disorder (n=49). Three factors were extracted for all groups on the basis of the scree plot. Eigenvalues were greater than 1 for all 3 factors for schizophrenia and schizoaffective disorder groups. For the nonschizophrenia and primary mood disorder groups, the eigenvalue for the third factor was 0.99, but a 3-factor solution yielded a better fit than a 2-factor solution.

The associations between individual domains and clinical variables were evaluated by analyses done in 2 different ways. In the first analysis, correlations between the clinical variables and severity score of each domain were assessed. Severity score of a domain was calculated by adding the raw scores of all symptoms contained in that domain. For example, the severity score of the positive domain was arrived at by adding the rating scale scores of "delusions" and "hallucinations" items. The possible ranges of severity scores for different domains are not equal, since the number of symptoms that compose each domain is different. This should not appreciably affect the findings, since the associations being studied are between each domain and an external variable and not across domains. Partial correlations were assessed for each domain, while controlling for the severity scores on the other 2 domains. Analysis was first done for the total sample (n=412) and then repeated for the subgroup of patients with schizophrenia (n=221) and for the subgroup with nonschizophrenia spectrum diagnoses (NSSD) (n=102). The NSSD subgroup excluded patients with schizophrenia, schizophreniform disorder, schizoaffective disorder, schizoid or schizotypal personality disorders, and organic mental disorders and the 2 patients with missing diagnoses.

In the second analysis, the 3 domains were evaluated to look for possible differences associated with the presence or absence of each individual domain. For this purpose, patients who definitely showed the presence of a domain (score of 2 or more on all the symptoms that compose the domain) were compared with patients who definitely did not show the presence of that domain (score of 0 on all the symptoms that compose that domain). Thus, patients who had only some of the symptoms of the domain as well as those with a score of 1 (questionable presence) on any of the symptoms of that domain were excluded from the analysis. This was done to minimize the possibility of results being confounded by the inclusion of equivocal cases. This method resulted in much smaller patient groups, and analysis could be carried out only for the total sample of patients and not for subgroups based on diagnostic categories. Two-tailed t test was used to look for differences in age at onset of first symptom, number of hospitalizations, frequency of hospitalizations, fraction of total illness duration spent in the hospital, Global Assessment Scale scores, and premorbid functioning. The χ² test was used to test differences in history of remissions and presence of tardive dyskinesia.

Results of the rotated factor solutions for patients with schizophrenia and all other diagnoses are shown in Table 3. Factor solutions for the subgroups with schizoaffective disorder and primary mood disorders are shown in Table 4. All 4 groups analyzed yielded fairly similar factor solutions, with the negative symptoms of flat affect, alogia, avolition, and asociality forming 1 factor (negative domain); delusions and hallucinations forming a second factor (positive domain); and formal thought disorder, disorganized behavior, and inappropriate affect accounting for the third factor (disorganization domain). To assess the dependency of this analysis on the specific variables included, the analysis was repeated after elimination of 1 variable at a time. Elimination of either 1 of the 2 variables that compose the positive domain resulted in the remaining variable combining with the disorganization domain to give a 2-factor solution. Elimination of any 1 of the other 7 variables did not significantly alter the 3-factor solution.

Of the 10 variables used to look for associations with individual psychopathological domains, 3 variables (age at first symptom, premorbid functioning, and history of remissions) showed a differential association with 1 or more psychopathological domains in both analyses. Results of the first analysis, showing the differential associations between these 3 variables and the severity score of each domain, are given in Table 5. Results of the second analysis, comparing patients who showed the definite presence of a domain with patients who did not show the presence of the domain, are given in Table 6. Results showed an earlier age at onset of symptoms to be associated with the disorganization and negative domains; poorer premorbid functioning with the negative domain; and a chronic course of illness with the disorganization and positive domains. Subgroup analyses (Table 5) showed that the associations between negative domain and premorbid functioning, and between disorganization domain and chronic illness course, were present in patients with schizophrenia as well as in the NSSD group.

Compared with the NSSD group, patients with schizophrenia had significantly higher mean severity scores on the disorganization (3.1 vs 4.2; P=.001), negative (4.7 vs 9.1; P<.001), and positive (4.2 vs 6.3; P<.001) domains. Comparison of the 2 groups also showed that patients with schizophrenia are significantly more likely to show the definitive presence of the domains (Table 7).

Results of the factor analysis showed a clear grouping of the symptoms into 3 factors for both schizophrenic and nonschizophrenic patients. Subgroup analyses of patients with primary mood disorders and schizoaffective disorders also disclosed a 3-factor pattern. These 3 factors are similar in their symptom composition to the factors reported by most other investigators using ratings from the SAPS and SANS administered to schizophrenic patients alone.^4-8 Our findings, based on a larger sample, confirm similar findings on a group of patients with familial bipolar disorder¹⁸ and on a heterogeneous group of psychotic patients.¹⁷ This supports the hypothesis that the 3 symptom domains identified in schizophrenia are not specific for that diagnostic category.

The results reported herein can be interpreted in different ways with respect to their implications for psychiatric nosology. One possibility is that common psychological or physiological pathways for the expression of psychosis are shared by schizophrenia and other disorders. This possibility would account for the phenomenological similarities and would argue that such similarities can be compatible with different diagnostic groups corresponding to distinct underlying disorders. The second possibility is that the phenomenological similarities reflect corresponding similarities in underlying disease. This would suggest that the same underlying disorders may be shared to varying degrees by different diagnostic groups.

Although the 3 symptom domains were identified across diagnostic categories, all 3 domains had significantly higher severity and prevalence in the schizophrenia group. This overrepresentation of the domains in schizophrenia is not an unexpected finding, since the domains were extracted from the SAPS and the SANS, which in turn focus particularly on the core features of schizophrenia.

Other investigators studying patients with schizophrenia found individual psychopathological domains to be independently associated with important clinical variables.^9-16 Results of this study showed individual domains to be differentially associated with age at onset of symptoms, premorbid functioning, and chronicity of illness course (Tables 5 and 6). Furthermore, these differential associations were similar in groups with and without schizophrenia. The finding reported herein, that individual domains show a differential association with important validating variables across diagnostic categories, more strongly supports the hypothesis that these domains are mutually independent and may arise from separate pathological processes. Further validation of this hypothesis would carry important implications for the concept of schizophrenia in particular and psychiatric nosology in general.

Despite multiple major and minor modifications to their symptom compositions, diagnostic groupings based on the categorical model remain etiologically uninformative. There is no clear demarcation between different diagnoses, and the neurobiological correlates of 1 diagnosis often correlate to varying degrees with those of other diagnoses. The Northwick Park functional psychosis study²² had shown that psychopharmacological agents treat specific symptoms or symptom groups rather than specific diagnostic categories. Dolan et al²³ compared patients with depression and schizophrenia for patterns of regional cerebral blood flow to the left dorsolateral prefrontal cortex. They found no difference between the 2 diagnostic groups, but found a significant difference when the total sample was regrouped on the basis of the presence or absence of poverty of speech. Taken together, these data suggest that many of the current diagnostic categories may arise from multiple and overlapping underlying disorders.

Discussing different ways of conceptualizing schizophrenia, Andreasen and Carpenter²⁴ suggested that "specific symptom clusters within schizophrenia [could reflect] different disease processes that combine in different ways in different patients." Such symptom clusters (or psychopathological domains) are different from subtypes, since they can, and often do, occur together in the same patient at the same time. Subtypes, on the other hand, are generally mutually exclusive. The findings of our study show that the domains of psychopathology are not specific for schizophrenia, and that the associations between domains and external validating variables can be seen across diagnostic categories. This suggests the possibility that each diagnostic category is composed of multiple symptom groups, with each group corresponding to a distinct disease process. Under this model, psychotic disorders result from a combination of discrete disease processes, each with a phenotypic presentation of a specific group of symptoms (domain), with such disease processes being non–mutually exclusive and occurring in different combinations in different patients. The current system of classification divides patients into diagnostic groups with the expectation that each group will correspond to an underlying disease process. The domains model proposed herein divides the symptoms within each patient into groups. This model is supported by the data reviewed in the preceding paragraph and by the results of this study. Further validation of the concept of psychopathological domains has potentially far-reaching consequences for the way psychotic disorders are conceptualized, classified, and treated.

This study has certain limitations that should be considered in interpreting the results. Since the study questions were not formulated at the time of data collection, the study is post hoc. Given the number of dependent variables studied, the possibility of chance findings should be considered. In this study, such a possibility may be offset by the fact that similar results were obtained by analyses done in 2 ways. Ratings on some of the variables, such as premorbid functioning and past Global Assessment Scale scores, are based on historical data and hence subject to limitations inherent to a retrospective design. Also, given the large sample, even small differences can be identified as being statistically significant. While such small differences may have limited clinical utility, they are still valuable for the purpose of hypothesis testing.

The 3 psychopathological domains identified were obtained from factor analysis of only 9 items included in the SAPS and SANS. This number may be considered small for factor analysis. Also, these symptoms account for only a fraction of the varied psychopathology seen in psychotic patients, or even in those diagnosed as having schizophrenia. Studies that used instruments other than the SAPS and SANS have suggested the existence of other domains, such as activation/excitation, insight or awareness of illness, and depression.^25-27

The utility of factor analysis depends on the precision with which the symptoms being measured are defined. The Maryland group has questioned the construct validity of SANS and proposed a trait-based approach (the deficit syndrome) to differentiate primary from secondary negative symptoms.^28-30 Symptoms that measure the same psychopathological phenomenon may lead to spurious correlations by aggregating together as a factor. Conversely, a symptom that is broadly defined may subsume more than 1 psychopathological phenomenon. In this study, delusions were found to load on both positive and disorganization domains in the schizoaffective disorders subgroup. It is possible that "delusion" as a symptom has been too broadly defined for our purposes. It has been suggested that delusions that are in the realm of thought content (eg, persecutory beliefs) are different from delusions that are more in the realm of the patient's subjective experience (eg, thought insertion).^17,31 As manifestations of psychopathology are defined with greater precision, the nature and content of the domains are likely to change.

Inability to find the disease processes underlying putative diagnostic categories has been a major factor hampering the progress of psychiatry during the past century. Such an inability, despite rapid technological advances in elucidating the structure and function of neural mechanisms, calls for alternative ways of conceptualizing psychiatric disorders. During the past 2 decades, beginning with the work of Strauss and colleagues,³² researchers have shown interest in studying the relative independence of different aspects of schizophrenia. Carpenter and colleagues³³ have argued for the usefulness of such a model in formulating unambiguous hypotheses that are amenable to falsification. Sufficient evidence has now accumulated to warrant investigating the applicability of such models to the nosology of all psychotic disorders. The domains of psychopathology model is a useful starting point for such an endeavor. Future research should focus on identifying and defining psychopathological domains with the help of statistical techniques, while parallel research evaluates the validity of such domains through their associations with biological and genetic mechanisms, longitudinal course, and response to treatment.

Accepted for publication February 18, 1997.

Supported in part by grants from the John D. and Catherine T. MacArthur Foundation, Chicago, Ill, and the National Alliance for Research on Schizophreniza and Depression, Chicago; and Developing Schizophrenia Clinical Research Center Grant OCPOMH50727A from the National Institute of Mental Health, Rockville, Md.

Presented in part at the annual meeting of the Society of Biological Psychiatry, New York, NY, May 2, 1996.

Reprints: Xavier F. Amador, PhD, 722 W 168th St, Unit 2, New York, NY 10032 (e-mail: XAI@Columbia.edu).