Brief Communication

Detection of Depression in Acute Schizophrenia: Sensitivity and Specificity of 2 Standard Observer Rating Scales
Matthias J Mller1, MD, Kay-Maria Mller2, Andreas Fellgiebel, MD3
Objective: To compare the psychometric properties of the Calgary Depression Rating Scale (CDRS) and the Hamilton Depression Rating Scale (HDRS) for severity assessment of depression in acute schizophrenia. Method: During clinical routine treatment, we investigated 119 inpatients with acute schizophrenia, using the CDRS, the HDRS, and a global 4-point Depression Severity Scale (DEP-SEV). We compared CDRS and HDRS sum scores regarding their diagnostic accuracy, with global severity of depression as the criterion. We estimated sensitivity and specificity on the basis of receiver operating characteristic curves. Results: According to global clinical ratings (DEP-SEV), 31% of patients had no depression, 19% had mild, 31% had moderate, and 19% had severe depression. Sensitivity was significantly higher (P < 0.05) for the CDRS than for the HDRS to assess mild (0.94 vs 0.76, cut-off 3 vs 10 points) or severe depression (1.00 vs 0.78, cut-off 11 vs 22 points); specificity was comparably high ($ 0.88) for both scales. Conclusion: Despite the fact that both scales were effective in separating mild, moderate, and severe depression, significant advantages emerged for the CDRS to detect mild or severe depression in schizophrenia. (Can J Psychiatry 2006;51:387392) Information on funding and support and author affiliations appears at the end of the article.

Clinical Implications
CDRS and HDRS scales are useful to grade depression severity in schizophrenia. The CDRS showed higher sensitivity in detecting mild and severe depression. Results can help clinicians to initiate psychosocial interventions and medication.

Limitations
All assessments of individual patients were carried out by the same rater. No fully standardized assessment of global depression severity was used. Replication is required because sampling bias cannot be entirely ruled out and sample size was restricted.

Key Words: depression in schizophrenia, Calgary Depression Rating Scale, Hamilton Depression Rating Scale, severity gradation

Can J Psychiatry, Vol 51, No 6, May 2006 W

387

The Canadian Journal of PsychiatryBrief Communication

epressive signs and symptoms are highly prevalent during all illness states of schizophrenia (1), and depression is often associated with a heightened risk for suicide, relapse or exacerbation, and lowered quality of life (2).

The study was approved by the local ethics committee and is in full accordance with the German law. Five experienced psychiatrists used the following tests during clinical routine: the semistructured, 9-item CDRS (911); the 17-item version of the HDRS (12); a global 4-point DEP-SEV (0 = no depression, 1 = mild, 2 = moderate, and 3 = severe depression); the PANSS (16); the severity item of the CGI (17); and other psychopathological scales not reported here. All raters had participated in continuous rater training. Intraclass correlation coefficients (5 raters and 20 schizophrenia patients) were $ 0.87 for CDRS, HDRS, and DEP-SEV ratings, indicating high interrater reliability. We assessed global severity of depression according to DSM-IV criteria of a major depressive episode with the following modifications: severity of depression was rated independently of psychotic features, and the time frame was set to 1 week. Means, SDs, and the correlation of depression scores are reported. We applied ROCs to evaluate and compare the diagnostic accuracy of the CDRS and HDRS sum scores, using global severity ratings as categorical criterion. We estimated AUC and standard errors, as well as sensitivity and specificity with corresponding 95%CIs (18), and statistically compared them between CDRS and HDRS scores (19). Nonoverlapping 95%CIs were regarded as statistically significant (P < 0.05).

After the advent of atypical antipsychotics, the optimism has increased that depressive symptoms could also become a target like positive and negative symptoms for more successful treatment of schizophrenia patients. Some data and recommendations corroborate this view (3,4), whereas results from several studies combining antidepressants and antipsychotics are still equivocal (5). Despite the broad agreement among psychiatrists that depression in schizophrenia is associated with considerable clinical burden (6,7), there is no consensus regarding the best treatment strategy or clear indication for initiating specific treatment (8). Thus the sensitive and specific assessment of depressive symptoms in schizophrenia is important for diagnostic, prognostic, and therapeutic reasons. The most appropriate dimensional assessment instrument today is the CDRS (911), but the HDRS (12), from which some modified items are also included in the CDRS (13), is still widely used. A score above 6 points on the CDRS has been proposed to separate schizophrenia patients with depression from those without depression (14); a further study has found optimal cut-off values of at least 7 points as related to major depression and CDRS scores of at least 4 points to detect minor depression in schizophrenia patients (15). However, studies directly comparing the accuracy of the CDRS and the HDRS for separating clinically graded mild, moderate, and severe depression in schizophrenia are still lacking.

Results
We investigated 119 inpatients (38% women and 71% paranoid subtype) with a mean age of 31.9 years, SD 10.7. Age at onset of schizophrenia was 26.8 years, SD 7.6; the inpatient treatment duration at the time of assessment was 19.3 days, SD 11.1. All patients were under neuroleptic treatment (72% atypical), and 28% received an antidepressant (19% a mood s ta b iliz e r , 1 1 % a n tich o lin e r g ic d r u g s , an d 5 5 % benzodiazepines) at the time of assessment. Mean PANSS total score was 77 points, SD 22; PANSS positive symptoms, 18 points, SD 8; and PANSS negative syndrome, 21 points, SD 8. The CGI severity score was 4.0 points, SD 1.3. Mean scores of depression assessment were as follows: CDRS 6.4, SD 5.8; HDRS 14.8, SD 8.7; and DEP-SEV 1.4, SD 1.1 points. The global assessment of depression severity ratings revealed that 31% of patients had no depression, 19% had mild depression, 31% had moderate depression, and 19% had severe depression. The distribution of CDRS and HDRS scores is shown in Figure 1, together with severity gradations. For the severity categories of depression (mild, moderate, and severe), the means and 95%CIs of CDRS and HDRS are shown in Figure 2. Generally, nonoverlapping CIs were yielded for CDRS, but not for HDRS, scores (only mild, compared with moderate, scores).
W Can J Psychiatry, Vol 51, No 6, May 2006

Methods
We investigated inpatients with stabilized acute schizophrenia diagnosed according to DSM-IV criteria. All patients gave their consent after the study procedure was fully explained.

Abbreviations used in the article

AUC CDRS CGI CI DEP-SEV HDRS NS PANSS ROC SD areas under the ROC curve Calgary Depression Rating Scale Clinical Global Impression confidence interval Depression Severity Scale Hamilton Depression Rating Scale not significant Positive and Negative Syndrome Scale receiver operating characteristic standard deviation

388

Detection of Depression in Acute Schizophrenia: Sensitivity and Specificity of 2 Standard Observer Rating Scales

Figure 1 Cumulative distribution of CDRS and HDRS scores

100% 90%
Cumulative proportion of patients

CDSS CDSS HDRS HAMD

>11 >22

"severe"

80% 70% 60% 50% 40% 30% 20% 10% 0% 0 5 10 15 20 Score 25 30 35 "none" >3 >10 >15 "mild" >7 "moderate"

n = 119 The classification none, mild, moderate, severe is based on the DEP-SEV.

Figure 2 Means of CDRS and HDRS scores

30 25 20 15 10 5 0
n = 119

Means (error bars: 95% CI for mean) no depression mild depression moderate depression severe depression

CDSS CDSS

HDRS HAMD

Can J Psychiatry, Vol 51, No 6, May 2006 W

389

The Canadian Journal of PsychiatryBrief Communication

Figure 3 Prediction of mild depression in schizophrenia by CDRS and HDRS scores

100

80

Sensitivity

60

40

20

CDSS CDSS HDRS HAMD

0 0
n = 119

20

40

60

80

100

100-Specificity

Table 1 Global severity gradation and prediction by CDRS and HDRS scores
Severity gradation (proportion of patients) Cut-off DEP-SEV 1 At least mild depression (69%) AUC sensitivity specificity Cut-off DEP-SEV 2 At least moderate depression (50%) AUC sensitivity specificity Cut-off DEP-SEV = 3 Severe depression (19%)
a b

CDRS 3 0.96 (0.02)

a b

HDRS 10 0.89 (0.03)a 0.76 (0.65 0.85)b 0.93 (0.80 0.98) 15 0.98 (0.02)a
b b b

Difference

P = 0.003 P < 0.05 ns

0.94 (0.86 0.98) 0.88 (0.73 0.96) 7 0.97 (0.02)a 0.88 (0.76 0.96) 11 0.99 (0.02)
a b

ns
b

0.90 (0.79 0.97) 22 0.93 (0.04)a

ns ns

0.97 (0.90 1.0)b

0.97 (0.90 1.0)b

AUC sensitivity specificity

ns P < 0.05 ns

1.0 (0.96 1.0)

0.79 (0.54 0.94)b 0.93 (0.86 0.97)b

0.92 (0.85 0.97)b

AUC (standard error) 95%CI

n = 119 Cut-off points are based on ROC analyses.

390

W Can J Psychiatry, Vol 51, No 6, May 2006

Detection of Depression in Acute Schizophrenia: Sensitivity and Specificity of 2 Standard Observer Rating Scales

HDRS scores were significantly correlated with CDRS scores (r = 0.78, P < 0.001), and both depression scales were significantly associated with DEP-SEV scores (CDRS r = 0.80, P < 0.001; HDRS r = 0.82, P < 0.001). Table 1 and Figure 3 show results of ROC analyses. Sensitivity ($ 0.76) and specificity ($ 0.88) were sufficiently high for both scales across all severity categories. Slight but statistically significant (P < 0.05) superiority emerged in favour of the CDRS for the assessment of at least mild depression and severe depression. When the conventional HDRS cut-off value for severe depression of 25 points was applied, sensitivity dropped to 0.63 (specificity 0.96).

References
1. Siris SG, Addington D, Azorin JM, Falloon IR, Gerlach J, Hirsch SR. Depression in schizophrenia: recognition and management in the USA. Schizophr Res 2001;47:18597. 2. Collaborative Working Group on Clinical Trial Evaluations. Atypical antipsychotics for treatment of depression in schizophrenia and affective disorders. J Clin Psychiatry 1998;59(Suppl 12):415. 3. Siris SG. Depression in schizophrenia: perspective in the era of atypical antipsychotic agents. Am J Psychiatry 2000;157:1379 89. 4. Mller HJ. Antidepressive effects of traditional and second generation antipsychotics: a review of the clinical data. Eur Arch Psychiatry Clin Neurosci 2005;255:8393. 5. Addington D, Addington J, Patten S, Remington G, Moamai J, Labelle A, and others. Double-blind, placebo-controlled comparison of the efficacy of sertraline as treatment for a major depressive episode in patients with remitted schizophrenia. J Clin Psychopharmacol 2002;22:20 5. 6. an der Heiden W, Konnecke R, Maurer K, Ropeter D, Hfner H. Depression in the long-term course of schizophrenia. Eur Arch Psychiatry Clin Neurosci 2005;255:174 84. 7. Kontaxakis V, Havaki-Kontaxaki B, Margariti M, Stamouli S, Kollias C, Christodoulou G. Suicidal ideation in inpatients with acute schizophrenia. Can J Psychiatry 2004;49:476 9. 8. Addington DD, Azorin JM, Falloon IR, Gerlach J, Hirsch SR, Siris SG. Clinical issues related to depression in schizophrenia: an international survey of psychiatrists. Acta Psychiatr Scand 2002;105:18995. 9. Addington D, Addington J, Maticka-Tyndale E, Joyce J. Reliability and validity of a depression rating scale for schizophrenics. Schizophr Res 1992;6:201 8. 10. Collins AA, Remington G, Coulter K, Birkett K. Depression in schizophrenia: a comparison of three measures. Schizophr Res 1996; 20:2059. 11. Mller MJ, Marx-Dannigkeit P, Schlsser R, Wetzel H, Addington D, Benkert O. The Calgary Depression Rating Scale for Schizophrenia: development and interrater reliability of a German version (CDRS-G). J Psychiatr Res 1999;33:43343. 12. Hamilton M. A rating scale for depression. J Neurol Neurosurg Psychiatry 1960;23:5662. 13. Addington D, Addington J, Schissel B. A depression rating scale for schizophrenics. Schizophr Res 1990;3:24751. 14. Addington D, Addington J, Maticka-Tyndale E. Specificity of the Calgary Depression Scale for schizophrenics. Schizophr Res 1994;11:23944. 15. Bressan RA, Chaves AC, Shirakawa I, de Jesus Mari J. Validity study of the Brazilian version of the Calgary Depression Scale for Schizophrenia. Schizophr Res 1998;22:419. 16. Kay SR, Fiszbein A, Opler LA. The Positive and Negative Syndrome Scale (PANSS) for Schizophrenia. Schizophr Bull 1987;13:26177. 17. National Institute of Mental Health. Clinical Global Impressions. In: Guy W, editor. EDCEU assessment manual for psychopharmacology. 2nd Rev Ed. Rockville (Maryland): NIMH; 1976. p 21722. 18. MedCalc Software. MedCalc Version 6.0. Mariakerke, (BEL): Med Calc; 2001. 19. Hanley JA, McNeil BJ. A method of comparing the areas under receiver operating characteristic curves derived from the same cases. Radiology 1983;148:83943. 20. Mller MJ, Himmerich H, Kienzle B, Szegedi A. Differentiating moderate and severe depression using the Montgomery-Csberg Depression Rating Scale (MADRS). J Affect Dis 2003;77:25560. 21. Mller MJ, Szegedi A, Wetzel H, Benkert O. Moderate and severe depression. Gradations for the Montgomery-Csberg Depression Rating Scale. J Affect Dis 2000;60:13740.

Discussion
This naturalistic clinical study replicated earlier findings that a large proportion of acute schizophrenia patients show signs and symptoms of depression with substantial severity along with coexisting positive, negative, and other schizophrenia symptoms. In the present study, we used a pragmatic global assessment of depression severity, partly following the DSM-IV criteria for major depression as a reference and comparable to the global assessment of illness severity on the CGI scale (17), which has become a worldwide standard. Nevertheless, this may be criticized because global and specific assessments of depression were carried out by the same rater and because the psychometric qualities of the reference assessment itself can be questioned. However, the lack of an established gold standard for the severity gradation of depression is also pertinent for studies in major depression (20,21). Although both depression scales were correlated and highly effective in separating mild, moderate, and severe depression, significant advantages emerged in favour of the CDRS. Use of the HDRS with conventional cut-off scores does not seem fully appropriate when applied to patients with schizophrenia. Thus the CDRS should be used with the presented cut-off values, particularly for a sensitive detection of mild (for example, subthreshold) depression at an early stage or to assess schizophrenia patients with severe depression who require thorough follow-up assessment and specific interventions (1,7).
Funding and Support This study received no commercial funding or support. Acknowledgments We thank Dr Bettina Kaufmann-Grebe, Dr Thomas Gesierich, and Dr Barbara Bggeln for their help in data collection.

Manuscript received September 2005, revised, and accepted January 2006. 1 Associate Professor, Department of Psychiatry, University of Mainz, Mainz, Germany; Acting Director, Clinic for Psychiatry and Psychotherapy Marburg-Sd, Marburg, Germany. 2 Psychiatrist, Department of Psychiatry, University of Mainz, Mainz, Germany. 3 Consultant and Senior Researcher, University of Mainz, Mainz, Germany. Address for correspondence: Dr MJ Mller, Clinic for Psychiatry and Psychotherapy Marburg-Sd, Cappeler Strasse 98, D-35039 Marburg, Germany; mjmueller@gmx.de

Can J Psychiatry, Vol 51, No 6, May 2006 W

391

The Canadian Journal of PsychiatryBrief Communication

Rsum : La dtection de la dpression dans la schizophrnie aigu : sensibilit et spcificit de 2 chelles dvaluation dobservateur rgulier
Objectif : Comparer les proprits psychomtriques de lchelle dvaluation de la dpression de Calgary (CDRS) et de lchelle dvaluation de la dpression dHamilton (HDRS) en ce qui concerne lvaluation de la gravit de la dpression dans la schizophrnie aigu. Mthode : Durant un traitement clinique rgulier, nous avons interrog 119 patients hospitaliss souffrant de schizophrnie aigu, laide de la CDRS, de lHDRS et dune chelle de la gravit gnrale de la dpression en 4 points (DEP-SEV). Nous avons compar les scores totaux la CDRS et lHDRS, lgard de lexactitude diagnostique, en utilisant le critre de la gravit gnrale de la dpression. Nous avons estim la sensibilit et la spcificit en fonction des courbes caractristiques actuelles du rcepteur. Rsultats : Selon lvaluation clinique gnrale (DEP-SEV), 31 % des patients navaient aucune dpression , 19 % avaient une lgre dpression , 31 % avaient une dpression modre et 19 % avaient une dpression grave . La sensibilit tait significativement plus leve (P < 0,05) pour la CDRS que lHDRS pour valuer la dpression lgre (0,94 compar avec 0,76, seuil dinclusion 3 et 10 points, respectivement) ou la dpression grave (1,00 compar avec 0,78, seuil dinclusion 11 et 22 points, respectivement); la spcificit tait comparativement leve ( 0,88) pour les deux chelles. Conclusion : Malgr le fait que les deux chelles taient efficaces pour distinguer la dpression lgre, modre et grave, la CDRS a dmontr des avantages significatifs pour dtecter la dpression lgre ou grave dans la schizophrnie.

392

W Can J Psychiatry, Vol 51, No 6, May 2006