OBJECTIVES:Validated diagnostic interviews are required to classify depression status and estimate prevalence of disorder, but screening tools are often used instead. We used individual participant data meta-analysis to compare prevalence based on standard Hospital Anxiety and Depression Scale - depression subscale (HADS-D) cutoffs of ≥8 and ≥11 versus Structured Clinical Interview for DSM (SCID) major depression and determined if an alternative HADS-D cutoff could more accurately estimate prevalence. METHODS:We searched Medline, Medline In-Process & Other Non-Indexed Citations via Ovid, PsycINFO, and Web of Science (inception-July 11, 2016) for studies comparing HADS-D scores to SCID major depression status. Pooled prevalence and pooled differences in prevalence for HADS-D cutoffs versus SCID major depression were estimated. RESULTS:6005 participants (689 SCID major depression cases) from 41 primary studies were included. Pooled prevalence was 24.5% (95% Confidence Interval (CI): 20.5%, 29.0%) for HADS-D ≥8, 10.7% (95% CI: 8.3%, 13.8%) for HADS-D ≥11, and 11.6% (95% CI: 9.2%, 14.6%) for SCID major depression. HADS-D ≥11 was closest to SCID major depression prevalence, but the 95% prediction interval for the difference that could be expected for HADS-D ≥11 versus SCID in a new study was -21.1% to 19.5%. CONCLUSIONS:HADS-D ≥8 substantially overestimates depression prevalence. Of all possible cutoff thresholds, HADS-D ≥11 was closest to the SCID, but there was substantial heterogeneity in the difference between HADS-D ≥11 and SCID-based estimates. HADS-D should not be used as a substitute for a validated diagnostic interview.

Background Depression symptom questionnaires are commonly used to assess symptom severity and as screening tools to identify patients who may have depression. They are not designed to ascertain diagnostic status and, based on published sensitivity and specificity estimates, would theoretically be expected to overestimate prevalence. Meta-analyses sometimes estimate depression prevalence based on primary studies that used screening tools or rating scales rather than validated diagnostic interviews. Our objectives were to determine classification methods used in primary studies included in depression prevalence meta-analyses, if pooled prevalence differs by primary study classification methods as would be predicted, whether meta-analysis abstracts accurately describe primary study classification methods, and how meta-analyses describe prevalence estimates in abstracts. Methods We searched PubMed (January 2008–December 2017) for meta-analyses that reported pooled depression prevalence in the abstract. For each meta-analysis, we included up to one pooled prevalence for each of three depression classification method categories: (1) diagnostic interviews only, (2) screening or rating tools, and (3) a combination of methods. Results In 69 included meta-analyses (81 prevalence estimates), eight prevalence estimates (10%) were based on diagnostic interviews, 36 (44%) on screening or rating tools, and 37 (46%) on combinations. Prevalence was 31% based on screening or rating tools, 22% for combinations, and 17% for diagnostic interviews. Among 2094 primary studies in 81 pooled prevalence estimates, 277 (13%) used validated diagnostic interviews, 1604 (77%) used screening or rating tools, and 213 (10%) used other methods (e.g., unstructured interviews, medical records). Classification methods pooled were accurately described in meta-analysis abstracts for 17 of 81 (21%) prevalence estimates. In 73 meta-analyses based on screening or rating tools or on combined methods, 52 (71%) described the prevalence as being for “depression” or “depressive disorders.” Results were similar for meta-analyses in journals with impact factor ≥ 10. Conclusions Most meta-analyses combined estimates from studies that used screening tools or rating scales instead of diagnostic interviews, did not disclose this in abstracts, and described the prevalence as being for “depression” or “depressive disorders ” even though disorders were not assessed. Users of meta-analyses of depression prevalence should be cautious when interpreting results because reported prevalence may exceed actual prevalence.

OBJECTIVE:Two previous individual participant data meta-analyses (IPDMAs) found that different diagnostic interviews classify different proportions of people as having major depression overall or by symptom levels. We compared the odds of major depression classification across diagnostic interviews among studies that administered the Depression subscale of the Hospital Anxiety and Depression Scale (HADS-D). METHODS:Data accrued for an IPDMA on HADS-D diagnostic accuracy were analysed. We fit binomial generalized linear mixed models to compare odds of major depression classification for the Structured Clinical Interview for DSM (SCID), Composite International Diagnostic Interview (CIDI), and Mini International Neuropsychiatric Interview (MINI), controlling for HADS-D scores and participant characteristics with and without an interaction term between interview and HADS-D scores. RESULTS:There were 15,856 participants (1942 [12%] with major depression) from 73 studies, including 15,335 (97%) non-psychiatric medical patients, 164 (1%) partners of medical patients, and 357 (2%) healthy adults. The MINI (27 studies, 7345 participants, 1066 major depression cases) classified participants as having major depression more often than the CIDI (10 studies, 3023 participants, 269 cases) (adjusted odds ratio [aOR] = 1.70 (0.84, 3.43)) and the semi-structured SCID (36 studies, 5488 participants, 607 cases) (aOR = 1.52 (1.01, 2.30)). The odds ratio for major depression classification with the CIDI was less likely to increase as HADS-D scores increased than for the SCID (interaction aOR = 0.92 (0.88, 0.96)). CONCLUSION:Compared to the SCID, the MINI may diagnose more participants as having major depression, and the CIDI may be less responsive to symptom severity.

Objectives Estimates of the prevalence of depression should be based on validated diagnostic interviews to determine case status and not depression screening tools or symptom rating scales, which are not intended for this purpose. Using cutoff scores from screening or rating scales to estimate prevalence tends to over-estimate prevalence substantially, particularly in low-prevalence groups (Thombs et al, CMAJ, 2018). Authors of meta-analyses, however, sometimes base estimates of depression prevalence on depression screening tools or rating scales. The objectives of this study were to (1) determine what depression ascertainment methods are commonly used to classify cases of depression in primary studies included in meta-analyses of depression prevalence; (2) determine what terminology is used in meta-analyses to describe prevalence rates based on different methods; and (3) examine the extent to which including studies that count positive screens as cases results in the overestimation of depression prevalence in recent meta-analyses of depression prevalence. Method We searched PubMed from 2008–2017 for meta-analyses that reported pooled prevalence of depression in the abstract. For each included meta-analysis, we recorded whether the abstract reported a pooled prevalence based on primary studies that used (1) diagnostic interviews only, (2) screening tools or rating scales only, and (3) a combination of diagnostic interviews, screening tools or rating scales, and other methods (e.g., chart diagnoses, self-report). If multiple prevalence estimates were reported for a category (e.g., for different subgroups), we extracted data only for the first prevalence estimate in the category. For each meta-analysis, for each prevalence estimate, we recorded whether the abstract indicated the types of ascertainment methods included, the terminology used to describe the prevalence value, and the number of studies pooled, the pooled sample size, and the pooled prevalence. For the combination category, we determined how many pooled primary studies used a validated diagnostic interview. Results 69 eligible articles were included, and 81 pooled prevalence estimates were extracted (9 for diagnostic interviews only, 36 for screening tools or rating scales only, 36 for combinations). Mean pooled prevalence was 17% for interviews only, 31% for screening tools or rating scales only, and 22% for combinations. Among 11 articles that reported prevalence for interviews or for combinations and also for screening tools or rating scales, prevalence was always higher based on screening tools or rating scales. Only 10 of 36 meta-analyses that combined studies that used screening tools or rating scales indicated this in the abstract; 22 of 36 referred to the prevalence as for ‘depression’ or a ‘depressive disorder’, despite using screening or rating tools. 5 studies that did not report a prevalence based on diagnostic interviews in the abstract provided one in the text; on average, it was half the value reported in the abstract. Conclusions Most meta-analyses of depression prevalence combine prevalence estimates from primary studies that used methods other than validated diagnostic interviews to assess depression. Some meta-analyses describe the resulting pooled prevalence as prevalence of ‘depressive symptoms’, but most describe prevalence of ‘depression’ or ‘depressive disorders’ despite not assessing disorders. Among 69 included meta-analyses, prevalence of depression was always higher when based on depression screening tools or rating scales than when based on diagnostic interviews or a combination of ascertainment methods. Including studies that assess depression based on methods other than validated diagnostic interviews exaggerates the true prevalence of depression. These are preliminary results. Final results will include, for the combination category, a comparison of published prevalence estimates and prevalence estimates based on re-analysis using only the studies that assessed depression using a validated diagnostic interview.