Improving the global identification of bipolar spectrum disorders: meta-analysis of the diagnostic accuracy of checklists

Shifting definitions and differences in the conceptualization of bipolar disorders have contributed to long diagnostic delays, poor reliability, and inconsistent findings. Rating scales are independent of clinical judgment and offer a reliable way to assess manic symptoms, making them good tools to improve both clinical and research diagnoses of bipolar disorder. However, there are dozens of candidates, with few obvious distinguishing characteristics, making it difficult to select one. Our goal was to metaanalyze the diagnostic accuracy of rating scales designed to identify [hypo]manic symptoms. Additionally, we explored potential moderator variables including global region, translation into a different language, and sample composition. Nearly 4000 articles were identified with searches in PubMed and PsycINFO, yielding 127 effect sizes from 103 studies that met the following inclusion criteria: (a) statistics reported by which a standardized effect size could be calculated, (b) participants age 18 + years, (c) reference diagnoses made by semistructured/structured diagnostic interview, (d) results published in English. Multivariate mixed regression models accounted for multiple effect sizes nested within sample. One hundred twenty-seven effect sizes across 14 rating scales were evaluated. There was significant heterogeneity across effect sizes; Cochran's Q(126 df) = 1622.08, p < .00005, and substantial variance components both within (σ2 = .057) and between samples (σ2 = .253). Four measures performed similarly well and significantly better than some competitors after controlling for design and reporting features. The best rating scales offer an inexpensive, efficient way to improve research and clinical diagnostic processes across diverse populations, and could also complement formal diagnoses for examining secular and cultural trends.