Screening for Depression in Children and Youth

Screening for depression among adolescents and children is controversial. According to the systematic review (SR) of screening instruments featured in this issue, there are good reasons for such scepticism. Roseman et al.¹ noted that the 20 articles that met their inclusion criteria (covering 17 studies and 20 tests) were not conducted in comparable settings, often used very small sample sizes, used weak determination/validation of cutoff scores, and did not exclude individuals who had an existing diagnosis (thus not needing to be screened). In line with their interpretation of this selected literature on test accuracy, Roseman et al. concluded that there is insufficient evidence to allow a recommendation, one way or the other, about the use of any of the instruments to detect major depressive disorder among children and adolescents.

It should be noted that three other recent reviews on this topic did make some cautious recommendations. However, Roseman et al.¹ built upon these reviews and used stringent protocols for their article search and identification, study selection, and evaluation of the likelihood of bias produced by the methods used in the articles that were reviewed. While no systematic review has been accused of perfection, Roseman et al. have provided a strong methodological case in support of their conclusions.

One might ask how a controversy about screening for depression among children and adolescents might arise. After all, many of us have been using the tests in question for years—presumably with some level of confidence (the review included well-known instruments such as the Beck Depression Inventory, the Children’s Depression Inventory, the Diagnostic Interview Schedule for Children, the Center for Epidemiological Studies Depression Scale, and several others). It is likely that the dynamic interaction of 1) an emphasis on evidence-based service delivery, 2) increases in knowledge due to the advent of electronic information systems, and 3) the rise of systematic review methodologies may have conspired in bringing questions to the fore about the adequacy of these, and other, health care activities. Whatever the reason, we need to focus on this surprising conclusion—that there is still much we need to know about our trusted screening instruments.

But what are the implications? First, simply put, the ability of the identified instruments to screen for child and adolescent depression has not been demonstrated by the studies that have been deemed to have best addressed the issue. Consequently, we cannot confidently use these tests for this purpose at this time. Perhaps we should thus ask whether this creates enough concern about their use in clinical settings to also merit a more stringent look at that domain.

Second, a very important matter is that it is not just that we are limited in what we can conclude about particular instruments but also that we cannot comfortably make pronouncements about the validity of screening programs, per se, since outcomes due to consequent program interventions will be confounded with matters of screening test accuracy.

Third, the serious conclusions of Roseman et al.¹ do not necessarily mean that the screening tests in question are not up to the job of identifying those with depression—what has been clearly demonstrated is that this class of instrument has not been adequately assessed. It follows, then, that it is too soon to abandon these tests or the notion of screening but rather to concentrate on the kind of study that can address the methodological and conceptual issues that have been outlined—presumably a very large and coordinated initiative. First, we need to be able to accurately screen cases, and only then can we see if the screening leads to a difference—not such an easy thing in itself.

Having said the above, we should perhaps consider that there are implications beyond the specific domain of child and adolescent depression. I have had recent experience with other SRs that have noted evidence shortfalls for screening instruments in somewhat different domains: one for mental health problems among seniors² and the other for evaluation of workplace productivity (which often involves depression).³ Additionally, an exploratory scan of the literature produced SRs detailing concerns about measurement instruments for anxiety among older adults⁴ and individuals with intellectual disabilities,⁵ as well as for violence risk among psychiatric patients.⁶ All of these reported some deficit in the literature existing at the time, ranging from insufficient data in support of the adoption of any test to a cautious recommendation for a few well-assessed tests that performed adequately. So it may be that mental health screening, or the screening tests themselves, may be questioned at a more widespread level.

Finally, a minority opinion (mine and a few others) is that it is not clear that prepubertal children actually exhibit major depression.⁷ Many of the instruments used for screening do not actually contain items that address depression, and child psychiatry caseloads generally do not contain many, if any, children who satisfy either of the two core criteria for DSM depression. But children do have disturbances that predict depression, irrespective of its name. In the end, it does not matter how future cases of depression are identified if it leads to proper attention. What might matter is how we conceptualize the changes that occur at the juncture of childhood and adolescence (presumably involving puberty). These are significant for most children, but they differ according to gender.⁸ Whatever one’s opinion on childhood depression, gender differences and the pre- to postpubertal makeup of testing samples need to be addressed more systematically.

In the end, Roseman et al.¹ have discussed the kinds of changes that would lead to better information on screening instruments, thus leading to the ability to evaluate the effectiveness of screening programs. It would seem that this would best involve a large-scale collaborative initiative that could be Canada-wide, allowing enough scope to 1) control for differences in settings parameters; 2) guarantee adequate sample size; 3) examine important individual differences such as gender, pubertal status, comorbidities (e.g., anxiety), and other significant factors; 4) develop national standards for comparable cutoff points; and 5) determine what an acceptable gold standard might be.