Design, Reporting, and Risk of Bias in Depression Screening Tool Diagnostic Accuracy Studies: A Series of Meta-research Reviews of Studies Published in 2018-2021
Bibliographic record
Abstract
Background: Depression accounts for more years lived with disability than any other medical condition.Depression screening has been proposed to identify individuals with unrecognized and untreated depression.A review of studies published in 2013-2015, however, found that a large proportion of primary studies of the diagnostic accuracy of depression screening tools were conducted in samples that inappropriately include individuals currently diagnosed or being treated for depression.This may lead to bias in estimates.Similarly, concerns have been raised regarding sample sizes of such studies.A review of studies published in 2013-2015 found that only 3% of studies reported sample size calculations, and, overall, sample sizes were too small to generate precise accuracy estimates.Finally, depression screening accuracy studies must be completely and transparently reported.No study, however, has evaluated the extent to which studies have reported results consistent with the Standards for Reporting of Diagnostic Accuracy Studies statement (STARD) reporting guideline.Methods: We searched MEDLINE (PubMed interface) on May 21, 2021 for primary studies of depression screening accuracy published January 1, 2018 or later.Through a series of 3 metaresearch reviews, we assessed (1) the proportion of studies that appropriately excluded individuals with a depression diagnosis or in treatment at the time of study enrolment (Study 1);(2) the proportion that reported sample size calculations, the proportion that reported confidence intervals (CIs), and precision, based on the width and lower bounds of 95% CIs for sensitivity and specificity (Study 2); and (3) adherence of studies to the STARD requirements (Study 3).Results: A total of 106 studies were identified and assessed.Only 18 studies (17%; 95% CI, 11% to 25%) appropriately excluded individuals with a depression diagnosis or in treatment at the time of study enrolment, which represented an improvement of 11% (95% CI, 3% to 20%) v compared to similar studies published between 2013 and 2015.Only 12 studies (11%) described a viable sample size calculation, which represented an improvement of 8% since the last review; 36 studies (34%) provided reasonably accurate CIs.Of 103 studies where 95% CIs were provided or could be calculated, 7 (7%) had sensitivity CI widths of 10%, whereas 58 (56%) had widths of 21%.Eighty-four studies (82%) had lower bounds of confidence intervals < 80% for sensitivity and 77 studies (75%) for specificity.These results were similar to those reported previously.Of 34 STARD items or sub-items, the number of adequately reported items per study ranged from 7 to 18 (mean = 11.5, standard deviation [SD] = 2.5; median = 11.5), and the number inadequately reported ranged from 3 to 17 (mean = 10.1,SD = 2.5; median = 10.0).There were 8 items adequately reported, 7 partially reported, 11 inadequately reported, and 4 not applicable in 50% of studies; the remaining 4 items had mixed reporting. Conclusion:Few depression screening accuracy studies appropriately excluded individuals already diagnosed or treated for depression; few studies reported sample size calculations, and sample sizes in most studies were too small to generate reasonably precise accuracy estimates.Appropriately designed studies, excluding individuals already diagnosed or treated for depression, are needed to generate realistic accuracy estimates that reflect what would be achieved in clinical practice.Future studies should conduct precision-based a priori sample size calculations to either attain desired precision levels or to understand limitations prior to initiating a study.Finally, recently published depression screening accuracy studies are not optimally reported.There is a need for attention to more fulsome reporting of methodological conduct of these studies.The research community,
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.033 | 0.152 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.003 | 0.000 |
| Bibliometrics | 0.002 | 0.004 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.002 |
| Open science | 0.002 | 0.002 |
| Research integrity | 0.000 | 0.001 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; both teacher heads agree on what is shown here.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".