Probability of Major Depression Classification Based on the SCID, CIDI, and MINI Diagnostic Interviews: A Synthesis of Three Individual Participant Data Meta-Analyses

Abstract

Introduction: Three previous individual participant data meta-analyses (IPDMAs) reported that, compared to the Structured Clinical Interview for the DSM (SCID), alternative reference standards, primarily the Composite International Diagnostic Interview (CIDI) and the Mini International Neuropsychiatric Interview (MINI), tended to misclassify major depression status, when controlling for depression symptom severity. However, there was an important lack of precision in the results. Objective: To compare the odds of the major depression classification based on the SCID, CIDI, and MINI. Methods: We included and standardized data from 3 IPDMA databases. For each IPDMA, separately, we fitted binomial generalized linear mixed models to compare the adjusted odds ratios (aORs) of major depression classification, controlling for symptom severity and characteristics of participants, and the interaction between interview and symptom severity. Next, we synthesized results using a DerSimonian-Laird random-effects meta-analysis. Results: In total, 69,405 participants (7,574 [11%] with major depression) from 212 studies were included. Controlling for symptom severity and participant characteristics, the MINI (74 studies; 25,749 participants) classified major depression more often than the SCID (108 studies; 21,953 participants; aOR 1.46; 95% confidence interval [CI] 1.11–1.92]). Classification odds for the CIDI (30 studies; 21,703 participants) and the SCID did not differ overall (aOR 1.19; 95% CI 0.79–1.75); however, as screening scores increased, the aOR increased less for the CIDI than the SCID (interaction aOR 0.64; 95% CI 0.52–0.80). Conclusions: Compared to the SCID, the MINI classified major depression more often. The odds of the depression classification with the CIDI increased less as symptom levels increased. Interpretation of research that uses diagnostic interviews to classify depression should consider the interview characteristics.

Accuracy of the Hospital Anxiety and Depression Scale Depression subscale (HADS-D) to screen for major depression: systematic review and individual participant data meta-analysis

“…39 Ideally, data from studies that used different reference standards could be combined but the different diagnostic characteristics of different types of interviews is a barrier. [30][31][32][33] Future studies are needed to develop different approaches to combine data even when different reference standards are used while capitalising on the substantial amount of individual participant data we have collected. Also, head-to-head studies for commonly used screening tools for depression are needed.…”

“…We found in previous IPDMAs that, compared with the semi-structured SCID interview, participants evaluated with the brief MINI were substantially more likely to be classified as having major depression. When the fully structured CIDI was used, participants with lower depressive symptom severity were more likely to be classified with major depression, but the opposite was true in those with greater symptom severity 30313233. Semi-structured interviews should be carried out by experienced diagnosticians and are considered to most closely replicate clinical diagnostic procedures 565758…”

“…When the fully structured CIDI was used, participants with lower depressive symptom severity were more likely to be classified with major depression, but the opposite was true in those with greater symptom severity. [30][31][32][33] Semi-structured interviews should be carried out by experienced diagnosticians and are considered to most closely replicate clinical diagnostic procedures. [56][57][58] For each reference standard category (semistructured, fully structured, and the MINI), for HADS-D cut-off values of 5-15 separately, bivariate random effects models were fitted from Gauss-Hermite quadrature.…”

“…Recent studies have shown that different diagnostic interview formats have substantively different likelihoods of classifying major depression. [30][31][32][33] An individual participant data meta-analysis (IPDMA) 34 involves a standard systematic review, followed by synthesis of line-by-line participant data from primary studies, rather than aggregated summary data. The advantages of an IPDMA with the HADS-D are the ability to include data from studies that administered the HADS-D and a diagnostic interview but did not publish accuracy results; to carry out subgroup analyses; to evaluate results excluding participants already diagnosed with or treated for depression who would not be screened in practice; and overcoming bias from selective cut-off reporting by including estimates of accuracy for all of the relevant cut-off values from the studies included in the meta-analysis.…”

“…Also, results were combined across reference standard diagnostic interviews, even though important differences in design and structure exist. Recent studies have shown that different diagnostic interview formats have substantively different likelihoods of classifying major depression 30313233…”