Potential Antidepressant Overtreatment Associated with Office Use of Brief Depression Symptom Measures

Discussion

In exploratory analyses of data from an RCT focused on participants with relatively few or no depression symptoms (defined by a PHQ-9 score <10), practice administration of a brief depression symptom measure during office visits was associated with hypothesized increases in depression diagnosis and antidepressant recommendation and/or prescription. Additional analyses suggested these findings were associated primarily with the PHQ-9 (not the PHQ-2).

Caution is required when interpreting our findings, given their preliminary nature and the limitations of the study design. The parent RCT did not include an expert diagnostic interview for depression. It is possible that some patients had clinical depression. However, PHQ-9 validation studies using an expert diagnostic interview reference standard indicate only about 1% of people scoring <10 have major depression,^4,6 the only form of depression consistently shown in RCTs to benefit from antidepressants.^13⇓–15 Also, we lacked information regarding the indication(s) for antidepressant recommendation and prescription. Thus we examined whether practice use of a brief measure was associated with the diagnosis of depression in the visit note to help gauge the likelihood that antidepressants were recommended or prescribed for depression versus other indications. That use of a brief depression symptom measure had parallel associations with both depression diagnosis and antidepressant recommendation and prescription (Table 3) suggests that a depression diagnosis was associated with these antidepressant recommendations and prescriptions.

In turn, our findings tentatively suggest practice administration of brief depression symptom measures may be associated with overdiagnosis of depression and overtreatment with antidepressants, both prevalent in primary care.^16,17 Our exploratory findings suggest the need for more definitive studies. The wide confidence intervals around the point estimates for the study outcomes stem from the relatively few participants with these events and with record evidence of use of a brief symptom measure, signaling further need for caution in interpreting the findings. Still, use of a brief symptom measure was associated with increased depression diagnosis and antidepressant recommendation and prescription in appropriately weighted analyses adjusted for potential confounders and in patient subsamples defined using progressively lower PHQ-9 cut points (with a correspondingly decreasing likelihood of major depression).^4,6 Thus one might expect similar odds of depression diagnosis and antidepressant recommendation and prescription associated with use of a brief symptom measure in unselected primary care samples.

While the observational nature of our analyses precludes causal inference, we consider 2 plausible hypotheses regarding possible mechanisms. One is that use of a brief symptom measure is simply a marker for providers or practices that place emphasis on identifying and treating depression. Prior work suggests considerable variation among providers in this regard.³¹ Highly engaged providers might be more likely to use brief symptom measures for case finding or screening. They might also be inclined to choose more comprehensive brief measures like the PHQ-9 over briefer options like the PHQ-2, potentially contributing to the differences in findings for these measures. At the same time, such providers might have a lower threshold for treating with antidepressants because of their greater interest in and sensitivity to depression cues, even for patients not meeting criteria for major depression.

A second possibility is that use of brief symptom measures could foster overtreatment with antidepressants by nonspecifically heightening provider consideration of major depression, a difficult diagnosis for many providers.³ Such prompting could occur directly (eg, provider is handed a completed brief measure), indirectly through patient activation (eg, patient asks provider questions about depression after working through a brief measure), or by both routes. Such nonspecific prompting might be triggered more by the detailed PHQ-9 than the briefer PHQ-2. However, both heightened provider engagement and nonspecific prompting associated with use of a brief symptom measure could lead to more appropriate treatment of patients with major depression. Studies designed to explore the balance of risks and benefits associated with the commonly used brief measures are needed.

The CTFPHC expressed concerns about overtreatment with antidepressants in their 2013 statement recommending against use of brief depression symptom measures.¹¹ Nonetheless, both the CTFPHC and the USPSTF acknowledged that no studies have examined the associations of the use of such measures with overtreatment with antidepressants.^7,11 Our findings call for RCTs designed and powered to address the associations of commonly used brief symptom measures with potential risks (eg, overtreatment with antidepressants leading to unnecessary adverse medication effects) and benefits (eg, increased recognition and treatment of depression, reduced functional limitations, fewer suicides) and to explore causal mechanisms.

In the absence of such RCTs, the USPSTF's qualified endorsement of office use of brief depression symptom measures was predicated on the presumption of benefits and did not consider possible risks.^7,10 The USPSTF acknowledged extrapolating from RCTs in which routine use of brief depression symptom measures was bundled with immediate enrollment of patients with major depression into collaborative care.⁷ These RCTs demonstrated improved outcomes for patients with major depression but apart from collaborative care did not examine the independent effects of use of a brief depression symptom measure. Confirmation in future studies of our findings suggesting potential overdiagnosis of depression and overtreatment with antidepressants associated with use of a brief symptom measure would suggest the need to reevaluate the USPSTF recommendation.

Future confirmation of our findings also would suggest the need to move beyond the use of simple brief depression symptom measures for case finding or screening to developing and studying the impact of novel tools that could help to better match depression treatment to patient need. Individually tailored patient activation computer programs that incorporate brief depression symptom measures represent one example.²² Rather than simply informing patients of their symptom measure scores, such programs provide individualized information and motivational messages.³² In the context of significant depression symptoms, the programs provide messages to activate the patient to discuss the symptoms with a provider and to enhance receptiveness to offers of treatment. Conversely, in the context of few or no depression symptoms, tailored programs inform patients that they are not likely to be depressed and not likely to benefit from treatment, potentially buffering against subsequent provider recommendation of an antidepressant. Initial RCT evidence indicates the promise of such tailored programs.²² Nonetheless, larger trials, powered to detect small but potentially clinically significant increases in diagnosis of depression and treatment with antidepressants in patients with few or no symptoms, are needed to determine whether the programs offer advantages over isolated use of brief symptom measures.

Beyond the study limitations noted previously, the parent RCT involved interventions to improve depression recognition and treatment, raising the possibility of a Hawthorne effect influencing our findings. Participating providers were not informed that depression was the focus of the study.^22,28 Nonetheless, some may have been prompted to consider and treat depression during participation (eg, after encountering patients activated by the interventions to discuss depression), potentially increasing their tendency to use brief measures and/or prescribe antidepressants. Also, we lacked information regarding the specifics of provider use of brief symptom measures, such as the indication (eg, case finding vs screening) and interpretation (eg, cut points used to define clinical depression). Despite these limitations, given there is little evidence supporting benefits of antidepressants among patients lacking major depression,^13⇓–15 the findings have potential clinical relevance regardless of the reasons for and approaches to using the brief symptom measures. If our key findings are replicated by others, studies designed specifically to address mechanisms will be required to tease out the “causes” of the observed association.

The accuracy of the medical record information we used is unknown, with uncertain net impact on the findings. Other ascertainment methods also have drawbacks. For example, audio or video recording of visits may alter patient and/or provider behaviors. The generalizability of the findings is also uncertain, particularly to categories of individuals ineligible for the parent RCT (eg, non-English speaking people, individuals with sensorimotor impairments precluding use of a touchscreen computer). Replicating our findings in studies with samples that include such people and that use other methods of ascertaining use of brief depression symptom measures, antidepressant recommendation and treatment, and their indications and applications will be useful.