Performance of the Bipolar Spectrum Diagnostic Scale in psychiatric outpatients

Abstract

Recent research has suggested that bipolar disorder, when defined to include milder variants such as bipolar II disorder and bipolar disorder not otherwise specified (NOS), is more prevalent than had been previously reported and often underrecognized. Recommendations for improving the detection of bipolar disorder have included careful clinical evaluations inquiring about a history of mania and hypomania and the use of screening questionnaires. The Bipolar Spectrum Diagnostic Scale (BSDS) was designed to be particularly sensitive to the milder variants of bipolar disorder. In the present report from the Rhode Island Methods to Improve Diagnostic Assessment and Services (MIDAS) project, we examined the operating characteristics of the BSDS in a large sample of psychiatric outpatients presenting for treatment. A total of 1,100 psychiatric outpatients were interviewed with the Structured Clinical Interview for DSM-IV and asked to complete the BSDS. Missing data on the BSDS reduced the sample size to 961, approximately 10% (n = 90) of whom were diagnosed with bipolar disorder. The sensitivity of the BSDS was similar for bipolar I disorder, bipolar II disorder, and bipolar disorder NOS/cyclothymia. A receiver operating curve (ROC) analysis indicated that cutoffs of 11 and 12 maximized the sum of sensitivity and specificity for the entire group of patients with bipolar disorder (area under curve = 0.80, p < 0.001). The cutoff point associated with 90% sensitivity for the entire sample of patients with bipolar disorder was 8. At this cutoff the specificity of the scale was 51.1% and positive predictive value was 16.0%. We compared the patients with and without bipolar disorder on each of the BSDS symptom items. The odds ratios were higher for the items assessing hypomanic/manic symptoms than items assessing depressive symptoms. We therefore examined the performance of a subscale composed only of the hypomania/mania items. The area under the curve in the ROC analysis was nearly identical to that of the entire scale (0.81, p < 0.001). With its high negative predictive value, the BSDS was excellent at ruling out a diagnosis of bipolar disorder; however, the low positive predictive value indicates that it is not good at ruling in the diagnosis. These data raise questions about the use of the BSDS as a screening measure in routine clinical psychiatric practice.