Despite established clinical diagnosis criteria for psychotic disorders, researchers from Harvard Medical School found a high variability of symptom presentation among dimensional- and categorial-defined diagnoses. These results were published in Psychiatric Research and Clinical Practice.

Inpatients (N=934) at McLean Hospital in Belmont, Massachusetts participated in this study. Patients were assessed by the Diagnostic and Statistical Manual of Mental Disorders (DSM)-IV Structural Clinical Interview (DSM-IV-TR), Young Mania Rating Scale (YMRS), Montgomery-Asberg Depression Rating Scale (MADRS), and Positive and Negative Symptom Scale (PANSS). Dimensional- and categorical-based diagnoses were compared between patients.

The study population consisted of 45% women aged mean 37.2 (standard deviation [SD], 12.9) years.

The investigators used a factor analysis to determine which items were most common between dimensional-based assessments (ie, YMRS, MADRS, and PANSS). The most parsimonious model was a 4-factor solution (manic, depressive, negative, and positive symptom factors) which best fitted RMSEA (threshold, 0.08) and SRMR (threshold, 0.05). PANSS items loaded on all factors, YMRS loaded on manic symptoms, and MADRS loaded on depressive symptoms.

Continue Reading

These 4-factor scores were compared with DSM-based diagnoses. Patients with schizophrenia or schizoaffective disorder scored highest among the positive and negative symptom factors, those with major depressive disorder had a higher depressive symptom factor, and those with bipolar disorder had high manic symptom and lower negative symptom factors.

Although patients with specific diagnoses tended to score higher for differing factors, a large proportion of overlap among all DSM diagnoses were observed. For example, when all pairs of diagnoses were compared, many factor-based scores best fit an alternate categorical-based diagnosis than the patient’s actual DSM diagnosis.

The investigators used a data-driven clustering approach to group patients together using their factor-based scores. A total of 7 clusters were identified which combined high and medium range scores. For example, 1 cluster consisted of individuals who had high depressive and negative symptoms with intermediate manic and positive symptoms scores and another cluster contained individuals who had intermediate manic symptom scores only. All clusters contained a mixture of patients who had schizophrenia, schizoaffective, and bipolar disorder.

The medication prescribed to patients correlated with differing diagnosis methods. Patients who had high positive symptom scores were more likely to be prescribed an antipsychotic (odds ratio [OR], 1.4; 95% CI, 1.2-1.6). Mood stabilizers were given to those who had high manic factor symptoms (OR, 1.5; 95% CI, 1.2-1.9) and a DSM diagnosis for bipolar disorder (OR vs schizophrenia, 9.7; 95% CI, 5.8-16.4), and antidepressants were most common among those with a low manic factor (OR, 0.4; 95% CI, 0-3-0.5), high depression factor (OR, 1.4; 95% CI, 1.3-1.6), and DSM diagnosis of major depressive disorder (OR vs schizophrenia, 2.0; 95% CI, 0.9-4.5).

The investigators chose to remove some factors (insight [YMRS], inner tension [MADRS], and hostility [PANSS]) from the analysis, which potentially limited this study.

These results supported the inclusion of both dimensional and categorical variables when making a diagnosis of psychosis. Further research of hybrid diagnostic models is needed.


Ravichandran C, Ongur D, Cohen BM. Clinical Features of Psychotic Disorders: Comparing Categorical and Dimensional Models. [published online October 9, 2020] Psych Res Clin Pract. doi: 10.1176/appi.prcp.20190053.