Improving Care for Depression: There's No Free Lunch FREE

Primary care clinicians know depression well. It steals success and satisfaction from their efforts to improve physical health and robs patients of joy, social networks, productivity, and stable lives. Yet despite their generally reasonable basic knowledge about depression and sense of responsibility for alleviating it (1 - 2), primary care clinicians continue to experience diagnostic and treatment failure (3). In answer to the pressing need to do better, the elegant study by Dobscha and colleagues in this issue (4) adds critical information to the existing body of work on improving depression outcomes in primary care, showing that in depression care, there's no free lunch.

Since the late 1980s, when researchers recognized that primary care is the major access point for depression care, studies have explored primary care–based interventions for improving depression outcomes. These studies showed that educational interventions alone, including reminders, did not improve depression outcomes (5 - 6). High-quality randomized trials, however, showed that organizational interventions based on systematic, standardized approaches, or care models, for managing depression in primary care practices can improve depression outcomes (7 - 8). Often termed “collaborative care for depression,” these care models feature an infrastructure for depression detection, assessment, triage to mental health specialty care, patient self-management, and treatment completion. The models often incorporate stepped care, meaning that more complex, treatment-resistant patients receive more intensive or specialized treatments (9). The core collaborative care resource is usually a depression care manager, commonly a registered nurse. The care manager fosters collaboration between patients, mental health specialists, and primary care clinicians and bridges the gaps in care through which hopeless, helpless, and side effect–prone patients with depression often fall.

Despite collaborative care model successes, few practices nationwide currently use this model routinely. In this sense, the model has failed—but not because it cannot be adopted by diverse practice settings. Successful dissemination requires only minimal direct contact between researchers and practices (10 - 12). Start-up support is required but at a level similar to that of other chronic illness care quality improvement efforts. With start-up support, small practices can implement care management by using office nurses (13) or by sharing a care management program (10 - 11).

Failure to disseminate collaborative care models is primarily due to difficulty in achieving financial stability for them, not for lack of cost-effectiveness but because of unreimbursed costs (14). Collaborative care models are as cost-effective as other commonly used outpatient treatments (15 - 19). Because depression treatment rates are low and major medical expenses (such as hospitalization) attributable to depression are rare, better depression care cannot generate enough short-term cost savings to the medical system to fully offset the expense of collaborative care. The potential economic benefits to medical care providers associated with collaborative care (15) are swallowed by the costs of treating more patients more thoroughly. Inadequate reimbursement is thus a major barrier to dissemination of collaborative care models (14).

In Dobscha and colleagues' study, researchers asked whether the collaborative care model could remain effective despite reducing the care manager's effort. To accomplish this, the model put more resources into primary care clinician education and decision support and fewer resources toward direct care manager support for patients, compared with successful programs in the past. For example, the care manager in Dobscha and colleagues' model had only 1 initial direct contact with patients, while the norm in previous studies was monthly contact for up to a year unless the patient was receiving mental health specialty care. Overall, the care manager in Dobscha and colleagues' study spent only 8 hours per week caring for 189 patients, compared with full-time effort for a similar number of patients in previous studies ((10), (20)).

The good news in Dobscha and colleagues' study is that the intervention changed provider behavior. Significantly more patients (10%) in the intervention group received appropriate antidepressants, and 10% more had their antidepressant treatment changed. Patients were more likely to be assessed for depression by their primary care clinicians (93% vs. 77%), and clinicians were more likely to have followed up with a depression-related action (85% vs. 54%). Significantly more patients (14%) received at least 1 mental health specialty visit.

Did these improvements in care affect outcomes? The bad news is that depression outcomes didn't change when compared with those for patients of control group clinicians. Depression symptoms at 6 or 12 months were similar, and the overall group means for the 2 depression symptom outcome measures remained in the range of major depression. In comparison, 2 large randomized trials of collaborative care used the same symptom outcome measure as that in Dobscha and colleagues' study and showed a 10% to 20% greater reduction in symptoms for intervention group patients. The final group mean depression scores were in the nondepressed range ((11), (20)).

Why didn't changes in care lead to better outcomes? Some explanations are easily countered. For example, perhaps the changes in process were not large enough to affect the outcomes. The improvements in antidepressant initiation in the study by Dobscha and colleagues, however, are at least as large as those in previous studies that showed effects ((10 - 11), (20)).

A second possible explanation is that studies in Veterans Affairs (VA) practices show smaller outcome effects from depression interventions. A multisite study of an effective program (20), however, included 2 VA sites, and these sites showed an outcome response that was equal to or greater than that at non-VA sites.

Could the patients in Dobscha and colleagues' study have been too chronically ill or complex to benefit from a primary care–based intervention? Study populations in other studies that did show improved depression care process outcomes had equal (11) or only slightly lower (20) baseline symptom severity. Many patients in these studies, like those in Dobscha and colleagues' study, had chronic disease, pain, posttraumatic stress disorder, and anxiety ((10 - 11), (20)).

These 3 explanations do not account for the different results of Dobscha and colleagues' study compared with those of previous studies. Two potentially important alternative explanations are the lack of care manager follow-up and incomplete stepped care in Dobscha and colleagues' model. The absence of care manager follow-up visits is the greatest difference from previous models and is most likely to have resulted in suboptimal outcomes. Frequent, ongoing patient support by care managers seems to be critical for many patients. In addition, if stepped care based on patient needs is fully applied, care managers proactively guide patients into more intensive treatments when needed, including psychotherapy. Increased access to psychotherapy was not a goal of Dobscha and colleagues' study, as in some other studies ((10), (20)), and evidently it did not occur, because the number of patients receiving more than 3 mental health specialty visits did not increase. The focus on depression care by primary care providers, rather than by mental health specialists, may have limited outcome improvements among some patients. Adding care manager follow-up and a stronger stepped care focus to the excellent provider change strategies used in Dobscha and colleagues' study might increase the effect on depression outcomes. Another refinement would be to determine which depressed patients benefit most from regular follow-up from a care manager, allocating the rest to a single contact.

Although Dobscha and colleagues' study is “negative” in one sense, its findings, in the context of previous work, can pave the way for the policy, research, and practice changes that are needed to ensure that patients receive high-quality, effective depression care (21). Currently, practices wishing to improve depression outcomes should adopt a collaborative care model that includes access to 6 to 12 months of care management. The alternative is to continue settling for low-quality depression care and its consequences: ongoing preventable damage to patients' lives. That outcome just isn't acceptable.

Lisa V. Rubenstein, MD, MSPH

Veterans Affairs Greater Los Angeles Healthcare System

RAND Health Program; David Geffen School of Medicine at the  University of California, Los Angeles Los Angeles, CA 91343