1 Lady Davis Institute for Medical Research, Jewish General Hospital, Montreal, Quebec, Canada; Department of Psychology, McGill University, Montreal, Quebec, Canada.
2 Library, Concordia University, Montreal, Quebec, Canada.
3 Lady Davis Institute for Medical Research, Jewish General Hospital, Montreal, Quebec, Canada; Department of Epidemiology, Biostatistics, and Occupational Health, McGill University, Montreal, Quebec, Canada.
4 Lady Davis Institute for Medical Research, Jewish General Hospital, Montreal, Quebec, Canada.
5 Library, McGill University, Montreal, Canada.
6 Lady Davis Institute for Medical Research, Jewish General Hospital, Montreal, Quebec, Canada; Department of Psychology, McGill University, Montreal, Quebec, Canada; Department of Epidemiology, Biostatistics, and Occupational Health, McGill University, Montreal, Quebec, Canada; Department of Medicine, McGill University, Montreal, Quebec, Canada; Department of Educational and Counselling Psychology, McGill University, Montreal, Quebec, Canada; Department of Psychiatry, McGill University, Montreal, Quebec, Canada. Electronic address: brett.thombs@mcgill.ca.

Objective: Database searches for studies of diagnostic test accuracy are notoriously difficult to filter, highly resource-intensive, and a potential barrier to quality evidence synthesis. We examined published meta-analyses of depression screening tool accuracy to evaluate the (1) proportion of included primary studies found in any online database in the original meta-analyses that were indexed in MEDLINE; (2) the proportion of patients from MEDLINE-indexed studies; and (3) the proportion of depression cases from studies indexed in MEDLINE.

Methods: MEDLINE and PsycINFO were searched from January 1, 2005 through October 31, 2014 for meta-analyses in any language on the accuracy of depression screening tools.

Results: We identified 16 eligible meta-analyses that included 398 primary study citations, which had been identified via an online database in the original meta-analyses, including 257 unique citations and 234 unique patient samples. The 234 unique patient samples included 69,957 total patients and 11,867 depression cases. Of these, 220 samples (94%) were from studies indexed in MEDLINE, including 97% of all patients and 96% of all depression cases. When applying a peer-reviewed search strategy in MEDLINE, 91% of all samples, 96% of patients and 95% of depression cases were retrieved. Results were similar for total and unique citations.

Conclusions: Restricting searches to MEDLINE may capture almost all eligible studies, patients and depression cases. Although not examined in the present study, MEDLINE may not be indexed as quickly as other databases. Thus, MEDLINE searches should be complemented by date-limited searches of other databases for recent citations.