You are on page 1of 8

Journal of Psychiatric Research 40 (2006) 2229


Detecting psychotic major depression using psychiatric rating scales

Jennifer Keller a,*, Rowena G. Gomez a, Heather A. Kenna a, Joel Poesner b, Charles DeBattista a, Benjamin Flores a, Alan F. Schatzberg a
b a Psychiatry and Behavioral Sciences, Stanford University School of Medicine, 401 Quarry Road, Stanford, CA 94305-5723, United States Department of Psychiatry, Washington University School of Medicine, 660 South Euclid Avenue, Box 8134, St. Louis, MO 63110, United States

Received 15 March 2005; received in revised form 27 June 2005; accepted 11 July 2005

Abstract Objective: The aim of this study was to assess whether individual or clusters of psychiatric symptoms can dierentiate patients with psychotic major depression (PMD) from those with nonpsychotic depression (NPMD). Method: Data were pooled from two studies investigating patients with moderate depression. A total of 129 subjects were studied. Patients in Sample 1 were unmedicated, while the majority of the patients in Sample 2 were taking psychotropic medications. Baseline rating scales were obtained for all subjects, including the Hamilton depression rating scale and the brief psychiatric rating scale (BPRS). We used discriminant function analyses, logistic regression, and ROC analyses to determine the patterns in symptoms that dierentiated the groups. Results: Psychotic patients were adequately dierentiated by the unusual thought content (UTC) item of the BPRS. Even mild UTC endorsement was an indicator of PMD. Furthermore, results suggest that the positive symptom subscale of the BPRS was even better at dierentiating PMD from NMPD patients. Sensitivity and specicity for this scale were 84% and 99%, respectively. Conclusion: Psychotic major depression is often undiagnosed and poorly treated. One reason for this trend is the failure of physicians to inquire in a more detailed manner about positive symptoms in patients with primary mood symptoms. Although physicians are not likely to have the time to conduct an entire BPRS during an evaluation, our results suggest that a few key symptoms, if assessed directly, may aid the psychiatrist to more eectively diagnose and subsequently treat their depressed patients. 2005 Elsevier Ltd. All rights reserved.
Keywords: Depression; Psychosis; Diagnosis; Psychotic major depression

1. Introduction Depression is one of the most common mental illnesses in the United States, with its prevalence estimated between 2.1% and 7.6% (Blazer et al., 1994; Weissman et al., 1996). Of those diagnosed with major depressive disorder (MDD), some patients also have psychotic symptoms (i.e., hallucinations or delusions). In a recent
Corresponding author. Tel.: +1 650 724 0070; fax: +1 650 723 8331. E-mail address: (J. Keller). 0022-3956/$ - see front matter 2005 Elsevier Ltd. All rights reserved. doi:10.1016/j.jpsychires.2005.07.003

study, Ohayon and Schatzberg (2002) reported that in the general population in ve European countries 2.4% of those surveyed met criteria for unipolar major depression, of whom, nearly 19% also had psychotic features. Thus, they found a prevalence of 0.4% of major depression with psychotic features. This percentage of major depressives with psychotic features is consistent with other estimates (Johnson et al., 1991). Although the DSM-IV suggests that the essential difference between major depression with psychotic features (PMD) and major depression without psychotic features (NMPD) is the presence of delusions or hallucinations, a

J. Keller et al. / Journal of Psychiatric Research 40 (2006) 2229


number of researchers have suggested that PMD and NPMD are very distinct syndromes, with dierent symptoms, clinical courses, familial patterns, cognitive abilities, and biological features (Rothschild, 2003; Schatzberg and Rothschild, 1992). Indeed, PMD patients often have longer duration of episodes (Coryell et al., 1987), a greater likelihood of recurrence of depression (Aronson et al., 1988; Lykouras et al., 1986), as well as a higher mortality rate than do NPMDs (Vythilingam et al., 2003). Recently, research has found that PMDs, as compared with NPMDs, have greater decits in various tests of cognition (Schatzberg et al., 2000). Frequently, the psychosis of someone with PMD is not as obvious as that seen in patients diagnosed with other psychotic disorders such as schizophrenia, which makes it more dicult to diagnosis. Indeed, Dubovsky (1991) concluded that about half of the depressed patients refractory to antidepressants have delusions and/or hallucinations of which the treating physician is unaware. The reasons why these symptoms are not reported vary, including that these symptoms are not bothersome to the patient, the patient does not want to be considered crazy, some deny the symptoms, and some physicians do not inquire about these specic symptoms. Furthermore, Parker et al. (1991) suggested that overt psychotic symptoms are not always seen in depression because they are masked by other prominent depressive symptoms, such as psychomotor disturbances. Even family members nd it dicult to recognize psychotic symptoms in some patients (Chambers et al., 1982). Often these patients do not seem ill enough to be psychotic. A recent case report of a Cushings disease patient with psychotic depression highlights the diculty in correctly making the diagnosis (Chu et al., 2001). A variety of methods are often used to assess psychiatric patients, and these methods have considerable impact upon diagnosis and treatment. Dawes et al. (1989) suggested that reliance on clinical judgment alone rather than statistical and standardized measures leads to less ` accurate diagnoses, which Leowe et al. (2004) recently conrmed. Many health professionals use unstructured interviews, although structured interviews and computer-based assessments are widely available. The time pressures for most mental health professionals often do not allow for thorough structured interviews or questionnaire, such as the SCID or Hamilton depression rating scale (HDRS). Miller et al. (2001) found that the structured and computer-based assessments had more diagnostic accuracy than the commonly used unstructured interview. Although there is no gold-standard for diagnosing in psychiatry, the SCID is wellestablished; however, the SCID can often be a long (two or more hours for a complete assessment) and cumbersome tool in clinical settings. Not only is the reliability and validity of an assessment tool important in the

clinical setting but so is the ease of administration. Given that major depression with psychotic features is often misdiagnosed, and therefore, mistreated, it is worth exploring if there are additional tools which can quickly assess specic symptoms to improve diagnosis and treatment in these patients. Psychotic major depression is typically severe in nature, and patients often have relatively high depression levels (Lykouras et al., 1986). Indeed, they often score higher on total depression scores than their NPMD counterparts (Coryell et al., 1984). Although severity of depression increases the likelihood of psychotic features. Ohayon and Schatzberg (2002) found that subjects with mild to moderate depression also commonly reported psychotic symptoms. Additionally, many severely depressed patients do not develop psychotic features (Endicott and Spitzer, 1979; Glassman and Roose, 1981). Thus, severity of depression alone does not entirely account for the presence of psychotic symptoms. Specic symptoms, however, appear to be more severe in PMD patients. For example, Rothschild et al. (1989) reported that while PMD patients had higher depression scores than NPMD, they found this was primarily due to elevations on the retardation and cognitive disturbance items in PMD patients. Researchers have consistently reported that more frequent and severe psychomotor diculties (either agitation or retardation) (Charney and Nelson, 1981; Lykouras et al., 1986) and increased feelings of guilt (Glassman and Roose, 1981; Lykouras et al., 1986; Parker et al., 1991) are associated with PMD. Beyond delusions and hallucinations, Parker et al. (1991) found that PMDs were distinct from NPMD melancholic patients on psychomotor disturbance, depressive content, diurnal variation, and constipation. Even when researchers have matched patients for total depression scores, PMD patients demonstrated higher scores on psychomotor disturbances (Glassman and Roose, 1981). A number of other symptoms have been reported to be greater in PMDs as compared to NPMDs, including depressed mood, paranoia, hypochondriasis, and anxiety. However, the empirical support for these is less robust and less consistent than are data supporting higher levels of psychomotor disturbances and increased guilt. Thus, it appears that although PMDs often have higher depression scores, it is likely due to specic, rather than a global, symptom elevation. Most of the research thus far has examined specic dierences in depressive symptomatology via the HDRS, without more systematic examination of psychotic symptoms. There are a number of scales and interviews used to assess the presence of psychotic symptoms. Our research group has debated how to most accurately assess psychotic symptoms within major depression, as there is no one scale that adequately


J. Keller et al. / Journal of Psychiatric Research 40 (2006) 2229

captures the intricacies of mood and the subtleties of the psychosis seen in major depression. In particular, the positive symptom subscale of the brief psychiatric rating scale (BPRS) is of interest, since hallucinations and delusions are the primary type of psychosis seen in MDD (Schatzberg and Rothschild, 1992). This subscale consists of conceptual disorganization, suspiciousness, hallucinations, and unusual thought content. Of these items, conceptual disorganization (i.e., thought disorder), the disruption of normal thought processes, is less typically seen in PMD patients. Suspiciousness, hallucinations, and unusual thought content (UTC) (delusions) are more common in PMD. Although specic depressive symptoms have been found to be associated with PMD, few studies have examined other psychiatric symptom patterns or predictors of PMD. In our research on psychotic major depression, our group has become increasingly interested in how best to assess PMD symptoms. We have routinely employed for many years the HDRS and the BPRS to assess symptoms. Given that major depression with psychotic features is often under diagnosed and therefore, mistreated, it is worth exploring if there are additional methods which can quickly and adequately assess specic symptoms to improve diagnosis and treatment in these patients. In this study, we investigated which item or group of items from these two clinician-administered measures best dierentiate between and successfully classify PMD and NPMD patients.

2. Method 2.1. Participants Data were pooled from participants in two separate studies, in which patients with PMD and NPMD were compared. A total of 129 participants were included in the present analyses (mean age = 40.87, SD = 13.4; 80% women, 84% Caucasian). Three additional patients had incomplete rating data and thus were not included in the data analyses. All participants were between 18 and 70 years of age, medically healthy, and had no drug or alcohol abuse or ECT in the previous six months. For all subjects, diagnoses were conrmed by the SCID interview in conjunction with the patients treating psychiatrist. After complete description of the study to the subjects, written informed consent was obtained. 2.2. Procedures Study 1. Twelve patients with PMD and 47 patients with NPMD were recruited from inpatient and outpatient psychiatric facilities at Stanford University Medi-

cal Center, Brigham and Womens Hospital, and Washington University Medical Center for a study on hypothalamicpituitaryadrenal axis (HPA) function in depression. At eligibility screening, a baseline psychiatric assessment was performed on all patients, including the structured clinical interview for DSMIII-R (SCID; Spitzer et al., 1990), portions of the schedule for aective disorders and Schizophrenia (SADS; Endicott and Spitzer, 1979), the 21-item HDRS (Hamilton, 1980), BPRS (Overall and Gorham, 1961), and the clinical global impression scale (CGI; Guy, 1976). Participants were required to be free of psychiatric medications for the two weeks preceding entry into the study. As part of the study, participants received a blinded challenge with intravenously administered oCRH, ACTH, hydrocortisone, or placebo after baseline in order to observe its eect on the HPA axis. Blood samples were collected over a 72-h period for determination of cortisol, ACTH, and monoamine levels. Subjects also underwent neuropsychological testing. Results of these procedures have been reported elsewhere (Belano et al., 2001; Schatzberg et al., 2000). Study 2. Thirty patients with PMD and 36 patients with NPMD were recruited at Stanford University Medical Center. Research sta conducted phone screens with interested individuals who responded to study advertisements. Individuals who met DSM-IV criteria for a current major depression episode and who met additional screening criteria were invited to participate in the study. Participants were given a baseline psychiatric assessment which included the same measures used in Study 1 (SCID, SADS, HDRS, BPRS, and CGI), except the newer version of the SCID (First et al., 1997) was used. Participants were allowed to remain on psychiatric medications, providing that doses were not altered for at least one week prior to the start of the study. 89% of PMDs and 63% of NPMDs were taking daily psychiatric medications. All participants also underwent functional MRI and neuropsychological testing, and blood samples were collected overnight for determination of cortisol and adrenocorticotropic hormone. These data will be reported elsewhere. In both studies, psychotic and nonpsychotic depressed patients were required to have a minimum score of 21 on the 21-item HDRS (Hamilton, 1980) and a minimum score of 7 on the Thase core endogenomorphic scale (Thase et al., 1983). The Thase scale includes: middle and late insomna, anhedonia, psychomotor retardation or agitation, weight loss, and diurnal variation. These two criteria were designed to ensure inclusion of participants with similar minimum levels of severity of endogenous-type symptoms. PMDs were also required to have a minimum of 5 on the positive symptom subscale of the BPRS (Overall and Gorham, 1961), which consists of four items: conceptual disorganization, suspiciousness,

J. Keller et al. / Journal of Psychiatric Research 40 (2006) 2229


hallucinations, and unusual thought content. All patients met DSM-IV criteria for a current major depressive episode, with or without psychotic features. We attempted to match patients on demographics, level of depression, and severity of endogenous symptoms, although this was not completely successful. Subjects from both studies were matched on age and education (see Table 1). Dierences in the demographic variables are likely due to changes in participant recruitment for Study 1 and Study 2. Study 1 only allowed unmedicated patients into the study, whereas in Study 2, patients were allowed to remain on psychotropic medications. It is likely that in order to be PMD and unmedicated, symptoms were less severe and patients were higher functioning. Because PMD is not simply a degree of severity of depression and psychotic symptoms can occur in mild-moderate depression as well (Ohayon and Schatzberg, 2002), data from the two studies were pooled in order to more accurately represent the spectrum of PMD. Importantly, similar results were found in all reported analyses whether

the studies were combined or when the Study 1 and Study 2 datasets were examined separately. 2.3. Study measures For the purposes of this paper, only the two clinician-administered rating scales, HDRS and BPRS, were examined. The HDRS evaluates depressed mood, vegetative, and cognitive symptoms of depression, and comorbid anxiety symptoms. The majority of items are rated from zero (absence of symptom) to four (most severe) but several items have a range from zero to two or three (most severe). The HDRS (21-item) is a widely used and accepted outcome measure for evaluating depression severity, with demonstrated acceptable levels of validity and reliability (Maier et al., 1988). The BPRS is an 18-item scale that evaluates positive symptoms, general psychopathology, and aective symptoms. Each item is rated from one (absence of symptom) to seven (extremely severe). The BPRS is also well-established with high validity and reliability across various psychiatric diagnoses (Hedlund and Vieweg, 1980). Within the BPRS is the positive symptom subscale, which consists of four items: conceptual disorganization, suspiciousness, hallucinations, and unusual thought content. 2.4. Data analyses Three major statistical techniques were conducted for this study two for hypothesis testing and one for exploration. These analyses included discriminant function analysis (DFA), binomial logistic regression (LR), and receiver operating characteristics (ROC) analysis. Below is a brief description of each of the analyses. DFA and LR were used to test specic hypotheses. DFA determines which of the independent variables in our study (i.e., the individual, non-overlapping items of the HDRS and BPRS) best distinguish the two naturally occurring groups, PMDs and NPMDs. We hypothesized that the positive symptoms items (i.e., unusual thought content, suspiciousness, hallucinations, and conceptual disorganization) would best predict group membership. This analysis provides standardized canonical discriminant function coecients which indicate the weight or degree of inuence that a predictor (independent) variable has on the dependent variable. It also includes only the signicant predictors in the nal analysis and excludes other extraneous predictors that do not aid in discriminating the clinical groups. More important, it classies cases into group membership (PMD or NPMD) and provides a measure of accuracy in classication given the signicant predictors. Similarly, binomial LR is a form of a regression analysis that predicts the dichotomous dependent variable (depression subtype) using the independent variables

Table 1 Demographic characteristics PMD (N = 12) Study 1 Age Educationa Gender Male Female Ethnicity Caucasian African-American Asian Latino 41.3 (12.8) 14.0 (2.2) 8 5 10 1 1 0 PMD (N = 37) Study 2 Age Educationb Gender Male Female Ethnicity Caucasian African-American Asian Latino 39.41 (12.2) 15.0 (2.8) 16 21 30 3 3 1 NPMD (N = 48) 42.7 (14.2) 14.7 (2.4) 17 33 43 2 1 2 NPMD (N = 32) 39.2 (12.5) 15.1 (2.1) 9 23 25 2 3 2 Analysis v2(3) = 21.7, p < 0.001 Analysis

F(1,59) = 0.088, ns F(1,58) = 0.794, ns v2(1) = 4.44, p < 0.05

F(1,68) = 0.006, ns F(1,65) = 0.331, ns v2(1) = 1.69, ns

v2(3) = 148.4, p < 0.001

Note: Means (SDs) are given for age and education in number of years. a Data missing for one subject. b Data missing for three subjects.


J. Keller et al. / Journal of Psychiatric Research 40 (2006) 2229

(predictors) and determines the amount of the dependent variables variance that the independent variables explain. It also shows the relative importance of predictors and can covary out any particular predictor or confounding variables. In this study, the LR model will control for signicant individual item predictor(s) that are identied via the DFA and will test whether a composite score (i.e., the BPRS positive symptom scale) provides additional explained variance and improves accuracy for classication of the PMD group. Exploratory analyses were also conducted by ROC analyses. Unlike DFA and LR which indicate the weight of each predictor, ROC indicates specic cut points of particular predictors that maximally discriminates the two groups. Specically, we used it to evaluate the discriminability of items from the HDRS and the BPRS and provide a cut-o score to determine whether a case is classied as PMD or NPMD. This statistical method was developed from signal-detection theory and has been used in biological and behavioral research (Erdreich and Lee, 1978) and is described in detail elsewhere (Kraemer, 1988). Briey, it permits calculation of overall test performance by considering sensitivity/specicity pairs for every possible threshold of a test. In this study, the tests are the individual, non-overlapping HDRS and BPRS items and the Positive Symptom subscale composite. The resulting ROC curve can be used for estimation of the optimal cut-o score according to the cost of false positive and false negative results.

Study 2 were signicantly more depressed and more psychotic than the PMDs from Study 1. The NPMDs from Study 2 also had signicantly higher total scores on the HDRS and BPRS. However, the NPMD samples did not signicantly dier on the positive symptom subscale. When the samples from both studies were combined, PMDs were more depressed and more psychotic than NPMDs on the psychiatric measures as expected. Thus, combining the samples from the two studies provides a wider range of depression and/or psychotic symptom severity for both clinical groups. Dierences were examined among the individual rating items between PMD and NPMD, with a Bonferroni correction for the number of comparisons. Consistent with the literature, PMD participants had signicantly higher ratings than NPMD patients on Guilt, t(127) = 4.96, p < 0.001, and psychomotor retardation, t(127) = 4.51, p < 0.001. Additionally, from the HDRS, depressed mood, suicidal ideation, work and activities, depersonalization, and obsessional symptoms items and from the BPRS, emotional withdrawal, conceptual disorganization, suspiciousness, hallucinations, uncooperativeness, unusual thought content, and blunted aect were all signicantly (all p < 0.001) higher in PMD subjects than NPMD. 3.2. Rating analyses Before the specied analyses were conducted, correlations between the individual HDRS and BPRS items were examined to determine if there was signicant overlap between any items. Seven items from the HDRS were excluded from the analyses because of high correlations with similar BPRS items, including depressed mood, guilt, psychomotor retardation, psychomotor agitation, anxiety, hypochondriasis, and paranoia. Correlations ranged from 0.59 to 0.89, with signicance

3. Results 3.1. Group descriptives Table 2 provides the descriptive data for the samples from the two studies on the rating scales. PMDs from

Table 2 Descriptive statistics on Hamilton depression scale total score, BPRS total score, and positive symptoms subscale from the BPRS Samples Study 1 (N = 12) M Psychotic major depression participants HDRS total 21.08 BPRS total 34.92 Positive symptoms 7.42 Study 1 (N = 48) M Non-psychotic major depression participants HDRS total 20.30 BPRS total 28.35 Positive symptoms 4.17 Note: Means (SDs) are given for each scale. * p < 0.05. ** p < 0.01. *** p < 0.001. (SD) (5.1) (4.57) (0.38) (SD) (5.4) (5.95) (2.61) Study 2 (N = 37) M 29.9 46.73 10.92 Study 2 (N = 32) M 23.34 33.53 4.38 (SD) (4.5) (4.5) (1.4) F(1,79) = 8.61** F(1,79) = 25.10*** F(1,79) = 0.926 (SD) (5.2) (7.8) (4.7) F(1,48) = 25.56*** F(1,48) = 23.20*** F(1,48) = 5.95* Test

J. Keller et al. / Journal of Psychiatric Research 40 (2006) 2229


p < 0.01 for all seven item pairs. Thus, in this analysis, only specic items from the HDRS and all the items from the BPRS were included. A discriminant analysis was then conducted to determine which items from the HDRS and BPRS best discriminate between PMDs and NPMDs. It was also used to classify cases into PMD or NPMD diagnostic groups and to indicate the accuracy of the classication given the signicant predictors. Using this statistical procedure, six items best distinguished group membership with 91% of the cases correctly classied (Wilks Lambda, F(6,127) = 35.1, p < 0.001). Sensitivity was 84% and specicity was 95% with these six items. The items that best discriminated from most to least important were usual thought content, obsessional symptoms, blunted aect, disorientation, hallucinations, and somatic complaints. Specically, the standardized canonical discriminant function coecients for these variables are as follows: unusual thought content, b = 0.643, obsessional symptoms = 0.475, blunted aect = 0.443, disorientation = 0.343, hallucinations = 0.294, and somatic complaints = 0.279. Based on these coecients, unusual thought content is the best variable that discriminates between PMDs and NPMDs. Next, an ROC analysis was conducted to explore whether unusual thought content is the best discriminating item for classifying the clinical groups and to provide the cuto(s) that will most accurately classify cases into diagnostic groups. In this analysis, the same individual items from the HDRS and BPRS that were used in the discriminant analysis were included. Results demonstrated that unusual thought content, with a cuto score of 2.0, best discriminated between the two clinical groups (j = 0.689, v2(1) = 64.00, p < 0.001). This indicates that any evidence of unusual thought content, even if very mild, suggests a diagnosis of PMD rather than NPMD. Those with a score of 1 indicate no evidence of any unusual thought content and are likely to be diagnosed as NPMD. Those with a score of 2 or more show evidence of Unusual Thought Content and are likely to be diagnosed as PMD. The cuto of 2 has a sensitivity of 69% and specicity of 96% (see Fig. 1). Once UTC has been considered, blunted aect signicantly helps to classify the remaining subjects into the two clinical groups (j = 0.473, v2(1) = 21.26, p < 0.001). Those with a score of less than 2 on unusual thought content and less than 3 on blunted aect are likely to be diagnosed as NPMD, with 93% correctly classied. This analysis supports the previous nding that unusual thought content is the best predictive item. A logistic regression was then conducted to determine if the BPRS positive symptom subscale would signicantly improve prediction between PMDs and NPMDs over and above unusual thought content. Unusual thought content was entered at Step 1 and the additional


ROC Curve






0.0 0.0 0.2 0.4 0.6 0.8 1.0

1 -Specificity
Fig. 1. Receiver operating characteristics (ROC) curves. Separation of psychotic major depression from nonpsychotic major depression (based on physician diagnosis) using the unusual thought content (UTC) item. A cut-o score of UTC = 2.0 reects an optimum of accuracy to separate PMD from NPMD patients, where sensitivity = 69% and specicity = 96%. The second ROC curve depicts the separation of psychotic major depression from nonpsychotic major depression (based on physician diagnosis) using the positive symptom subscale (PSS) from the BPRS. A cut-o score of positive symptom subscale = 6.0 reects an optimum of accuracy to distinguish PMD from NPMD patients, where sensitivity = 84% and specicity = 99%.

three BPRS positive symptom subscale items were entered at Step 2 as the independent variables. At Step 1, unusual thought content explained 44% of the variance (Cox and Snell R2) and correctly classied 86% of the sample, v2(1) = 75.04, p < 0.001. At Step 2, the positive symptom subscale (less UTC) explained an additional 8% of the variance and correctly classied 92% of the sample, v2(1) = 18.79, p < 0.001. The additional other items from the positive symptom subscale (hallucinations, suspiciousness, and conceptual disorganization) provided additional predictive accuracy after UTC was considered. Lastly, an ROC analysis was conducted with the positive symptom subscale. Our goal was to explore how well this scale classies the diagnostic groups and to provide a cuto score that would produce the most accurate classication. The positive symptom scale was quite eective for classifying PMD or NPMD patients (j = 0.848, v2(1) = 94.02, p < 0.001). Its sensitivity was 84% and specicity was 99%, with a cuto score of 6. This cuto score of 6 indicates that scoring very mild on any two of the positive psychotic symptoms or scoring at least in the mild range (3 or higher) on any one psychotic symptom would likely result in a diagnosis of psychotic major depression.


J. Keller et al. / Journal of Psychiatric Research 40 (2006) 2229

4. Discussion The present ndings replicate previous studies on psychotic major depression. Consistent with the literature, a number of individual depression items from the HDRS dierentiated the PMD and NPMD groups. Specically, retardation and guilt scores were higher in the PMD group. A number of other depression items were higher in the PMD group, including depressed mood, suicidality, work and activities, depersonalization, and obsessional symptoms. Although the literature is inconsistent, the rst three items in the preceding list are sometimes found to be higher among PMD patients. An important feature of this study is that it combined a wide-range of depression severity, both for the PMD and NPMD groups. Patients from Study 1 were unmedicated and experienced less severe depression than those in Study 2. Additionally, we attempted to match NPMD and PMD subjects on minimum levels of depression and endogenous symptoms. Thus, our results are not just based on depression severity. Our study goes beyond the previous research that differentiated PMDs from NPMDs using individual items from depression rating scales. We examined patterns of response on the HDRS, as well as the BPRS, to aid our understanding and denition of PMD. The present results suggest that inquiring about particular items that make up the Positive Symptom scale can assist in making a more accurate dierential diagnosis of PMD or NPMD. Although the specicity is not 100% when compared to the SCID, it could improve the rate at which psychotic depressed patients are currently identied. Specically, UTC, with a cuto score of 2.0, appears to best discriminate between PMD and NPMD. UTC directly assesses a patients unusual, odd, strange, or bizarre ideas and beliefs. It is not, however, an index of thought disorder, as delusional or paranoid patients may present their odd beliefs in a very clear manner. Importantly, endorsement of even very mild severity on this item suggests that the patient would be far more likely to be diagnosed with PMD. Although other researchers have found that psychomotor disturbances dierentiated melancholic psychotic patients, this factor was not a discriminating factor in the discriminant function analysis. This is not to say that psychomotor disturbances are not important symptoms. Indeed, psychomotor disturbance was dierent between groups even though the eligibility criteria were designed to minimize dierences in endogenous symptoms across groups. Our results do suggest that psychomotor retardation does not provide better discrimination over and above that obtained for the presence of delusional thinking. In addition to queries about delusional thinking, the entire BPRS positive symptom scale is an even better tool to dierentiate PMD from NMPD patients, with

a sensitivity of 84% and specicity of 99% in our sample. A cuto score of 6 on this scale best dierentiates the patient groups. A more detailed evaluation of positive symptoms would entail further assessment about suspiciousness and hallucinations. Conceptual disorganization is also included in this subscale, although it is less commonly seen in PMD patients. The extensive disorganization of speech that is sometimes seen in schizophrenia or schizoaective disorder patients is typically not seen in PMD. Furthermore, conceptual disorganization is a more overt psychotic symptom, and therefore, psychosis is easier to detect when thought disorder is present. Our study addressed how best to detect psychosis when it is less obvious, such as in PMDs. Inquiring about hallucinations and suspiciousness, in addition to unusual thought content, increases the likelihood of accurately detecting psychotic features in major depression. Furthermore, inquiring about and detecting these psychotic features can lead to more accurate diagnoses and more appropriate treatment. The prevalence of major depression is between 2.1% and 7.6%, while major depression with psychotic features is estimated to be about 0.4% (Ohayon and Schatzberg, 2002). Although the prevalence of PMD is relatively small, Dubovsky and Thomas (1992) point out that psychotic features are frequently present in treatment-resistant depressed patients and these symptoms are frequently missed by treating physicians. Physicians likely will not have the time to conduct an entire SCID (which is reliable and well-known, albeit lengthy) or even an entire BPRS during an evaluation; however, our results suggest that a few key symptoms, particularly delusional thinking, when assessed directly, may aid practitioners in more precise diagnosis and subsequent improved treatment of their depressed patients.

Acknowledgment This study MH50604. was supported by NIMH Grant

Aronson TA, Shukla S, Gujavarty K, Ho A, DiBuono M, Khan E. Relapse in delusional depression: a retrospective study of the course of treatment. Compr Psychiatry 1988;29(1):1221. Belano JK, Kalehzan M, Sund B, Fleming Ficek SK, Schatzberg AF. Cortisol activity and cognitive changes in psychotic major depression. Am J Psychiatry 2001;158(10):16126. Blazer DG, Kessler RC, McGonagle KA, Swartz MS. The prevalence and distribution of major depression in a national community sample: the National Comorbidity Survey. Am J Psychiatry 1994;151(7):97986. Chambers WJ, Puig-Antich J, Tabrizi MA, Davies M. Psychotic symptoms in prepubertal major depressive disorder. Arch Gen Psychiatry 1982;39(8):9217.

J. Keller et al. / Journal of Psychiatric Research 40 (2006) 2229 Charney DS, Nelson JC. Delusional and nondelusional unipolar depression: further evidence for distinct subtypes. Am J Psychiatry 1981;138(3):32833. Chu JW, Matthias DF, Belano J, Schatzberg A, Homan AR, Feldman D. Successful long-term treatment of refractory Cushings disease with high-dose mifepristone (RU 486). J Clin Endocrinol Metab 2001;86(8):356873. Coryell W, Endicott J, Keller M. The importance of psychotic features to major depression: course and outcome during a 2-year followup. Acta Psychiatr Scand 1987;75(1):7885. Coryell W, Pfohl B, Zimmerman M. The clinical and neuroendocrine features of psychotic depression. J Nerv Ment Dis 1984;172(9):5218. Dawes RM, Faust D, Meehl PE. Clinical versus actuarial judgment. Science 1989;243(4899):166874. Dubovsky SL. What we dont know about psychotic depression. Biol Psychiatry 1991;30(6):5336. Dubovsky SL, Thomas M. Psychotic depression: advances in conceptualization and treatment. Hosp Community Psychiatry 1992;43(12):118998. Endicott J, Spitzer RL. Use of the research diagnostic criteria and the schedule for aective disorders and schizophrenia to study aective disorders. Am J Psychiatry 1979;136(1):526. Erdreich LS, Lee ET. A diagnostic interview: the schedule for aective disorders and schizophrenia. Arch Gen Psychiatry 1978;35(7):83744. First MB, Spitzer RL, Gibbon M, Williams JBW. Structured clinical interview for DSM-IV-TR Axis I disorders. Patient Edition (SCIDI/P, 2/2001 revision). Washington, DC: American Psychiatric Press; 1997. Glassman AH, Roose SP. Delusional depression: a distinct clinical entity? Arch Gen Psychiatry 1981;138:8313. Guy W. Early Clinical Drug Evaluation Unit (ECDEU) assessment manual for psychopharmacology. Revised. Bethesda, MD: NIMH; 1976. Hamilton M. A rating scale for depression. J Neurol, Neurosurg Psychiatry 1980;23:4965. Hedlund JL, Vieweg BW. The brief psychiatric rating scale: a comprehensive review. J Operational Psychiatry 1980;11:4965. Johnson J, Horwath E, Weissman MM. The validity of major depression with psychotic features based on a community study. Arch Gen Psychiatry 1991;48(12):107581. Kraemer HC. Assessment of 2 ;2 associations: generalization of signal detection methodology. Am Statistician 1988;42:3749. ` ` Leowe B, Spitzer RL, Greafe K, Kroenke K, Quenter A, Zipfel S, et al.. Comparative validity of three screening questionnaires for


DSM-IV depressive disorders and physicians diagnoses. J Aect Disord 2004;78(2):13140. Lykouras E, Malliaras D, Christodoulou GN, Papakostat Y, Voulgari A, Tzonou A, et al.. Delusional depression: phenomenology and response to treatment: a prospective study. Acta Psychiatr Scand 1986;73:3249. Maier W, Philipp M, Heuser I, Schlegel S, Buller R, Wetzel H. Improving depression severity assessment I. Reliability, internal validity and sensitivity to change of three observer depression scales. J Psychiatr Res 1988;22(1):312. Miller PR, Dasher R, Collins R, Griths P, Brown F. Inpatient diagnostic assessments: 1 Accuracy of structured vs unstructured interviews. Psychiatry Res 2001;105(3):25564. Ohayon MM, Schatzberg AF. Prevalence of depressive episodes with psychotic features in the general population. Am J Psychiatry 2002;159(11):185561. Overall JE, Gorham DE. The brief psychiatric rating scale. Psychological Rep 1961;10:799812. Parker G, Hadzi-Pavlovic D, Hickie I, Boyce P, Mitchell P, Wilhelm K, et al.. Distinguishing psychotic and non-psychotic melancholia. J Aect Disord 1991;22(3):13548. Rothschild AJ. Challenges in the treatment of depression with psychotic features. Biol Psychiatry 2003;53(8):68090. Rothschild AJ, Benes F, Hebben N, Woods B, Luciana M, Bakanas E, et al.. Relationships between brain CT scan ndings and cortisol in psychotic and nonpsychotic depressed patients. Biol Psychiatry 1989;26(6):56575. Schatzberg AF, Posener JA, DeBattista C, Kalehzan BM, Rothschild AJ, Shear PK. Neuropsychological decits in psychotic versus nonpsychotic major depression and no mental illness. Am J Psychiatry 2000;157:1095100. Schatzberg AF, Rothschild AJ. Psychotic (delusional) major depression: should it be included as a distinct syndrome in DSM-IV? Am J Psychiatry 1992;149(6):73345. Spitzer RL, Williams JB, Gibbon M, First MB. Structured Clinical Interview for DSM-III-R Patient Edition. Washington, DC: American Psychiatric Press; 1990. Thase ME, Hersen M, Bellack AS, Himmelhoch JM, Kupfer DJ. Validation of a hamilton subscale for endogenomorphic depression. Journal of Aective Disorders 1983;5:26778. Vythilingam M, Chen J, Bremner JD, Mazure CM, Maciejewski PK, Nelson JC. Psychotic depression and mortality. Am J Psychiatry 2003;160(3):5746. Weissman MM, Bland RC, Canino GJ, Faravelli C, Greenwald S, Hwu H-G, et al.. Cross-national epidemiology of major depression and bipolar disorder. JAMA 1996;276(4):2939.