Late-life depression frequently differs from early-life depression in its clinical characteristics, particularly if it is late in onset or accompanied by signs of executive dysfunction or vascular disease. Late-life depression is often associated with executive dysfunction (25–29), a neuropsychological expression of frontal system impairment, with a clinical presentation of depression resembling medial frontal lobe syndromes (25). Depression affects cognitive function in all age groups, but the executive tasks of response inhibition and sustained effort are more frequently impaired in geriatric depression (30). Executive dysfunction generally subsides as depression improves, but tends to persist after remission of depression (31–36). When depressed elderly patients have executive dysfunction, they are more likely to have reduced interest in activities, more profound psychomotor retardation (25), and poor and unstable response to antidepressants (35,37–39).
The clinical presentation of late-onset geriatric depression with comorbid vascular disease is similar to that of geriatric depression with executive dysfunction. Compared to elderly patients with early-onset depression and no vascular risk factors, patients with late-onset major depression and vascular risk factors have shown greater impairment in frontal functions, poorer insight, more psychomotor retardation, less agitation and guilt, and more disability (40,41). A comparison of magnetic resonance imaging (MRI)-defined vascular and non-vascular depression showed that the vascular group had significantly greater age, age of onset, anhedonia, and disability, but less psychosis. Despite some negative findings (42–44), much past and recent evidence indicates that vascular depression predicts poorer response to antidepressants (10,38,43,45–49).
Family history of depression is less common in patients with late-onset depression than in elderly patients with early-onset recurrent depression (50,51); and less common in “vascular depression” than in non-vascular depression (52,53). In a recent, large Swedish twin-pair study, history of early-onset depression in one member of a twin pair was associated with high lifetime risk of depression in the other member. In contrast, late-onset depression was associated with high cotwin risk of vascular disease (54).
Finding genes predisposing to depression has been a formidable task. Depression has polygenic inheritance, thus making it difficult to identify the contribution of individual genes. To overcome this obstacle, research increasingly focuses on genes related to specific behavioral or biological functions (endophenotypes) related to depression.
Genetic studies of the serotonin transporter exemplify this work. The serotonin transporter is the site of action for serotonin reuptake inhibitors (SSRIs). A polymorphism of the serotonin transporter gene promoter region (5-HTTLPR) involves a 44-base pair insertion (L allele) or deletion (S allele). The S allele has been shown to reduce gene expression, thus reducing serotonin reuptake (55).
In addition to studies relating the S allele to risk for depression (56–58), a number of studies have found an association between the S allele and increased risk of vascular disease.
Elevated blood cholesterol and triglycerides, heart disease, and myocardial infarction have been more common among S allele carriers than L allele homozygotes (59); and after acute myocardial infarction, depressive symptoms and negative cardiac outcomes including cardiac death were more common in S-carriers than L-homozygotes (60). In one of our own recent late-life depression studies, we found that, compared with L-homozygotes, S-carriers had microstructural white matter abnormalities (lower fractional anisotropy, to be explained below) in frontolimbic brain regions as well as a lower remission rate of depression (61).
Neuroradiological and histopathological studies have found associations among depression, executive dysfunction, and brain abnormalities, most notably those affecting the structural integrity of frontostriatal circuits, which include subcortical regions. Executive dysfunction and depression are hypothesized to be related to fronto-striato-limbic network abnormalities (3). Five such frontostriatal circuits have been described (62,63). Glutamate, enkephalins, and gamma-aminobutyric acid (GABA) are important neurotransmitters in these circuits, with acetylcholine and dopamine serving a modulating role. Because these circuits appear to mediate positive affect-guided anticipation, damage to them, resulting in failure to anticipate incentives, is hypothesized to be a mechanism leading to depression (40).
Macroscopic and microscopic changes to brain regions have been associated with mood disorders in histopathological and neuroimaging studies. In post-mortem studies of depressed patients, glia reduction has been observed in the subgenual prelimbic anterior cingulate gyrus (64). Bipolar and unipolar depression studies have reported neuron abnormalities in the dorsolateral prefrontal cortex (65). Radiological studies have shown low orbitofrontal (66,67), anterior cingulate (68), and hippocampal volumes (69,70) in depressed elderly patients compared to healthy elderly controls, while reports of amygdala volumes have differed (71). Studies of white matter integrity in normal aging have shown a greater tendency for decline in prefrontal white matter with advancing age compared with other areas of the brain (72). Prefrontal white matter microstructural abnormalities have correlated with poorer performance on tasks of executive function, and have been hypothesized to reflect a disconnection state that can increase the risk of geriatric depression (73).
Depression has been associated with cerebral infarcts, even when no obvious neurological symptoms are present. In a Dutch study, such “silent” infarcts occurred five times more frequently than cerebral infarcts with peripheral neurologic signs (74). In a large US study, 28% of elderly subjects with no previous history of transient ischemic attack or stroke had evidence of previous infarcts. Eighty-one percent of these had lacunes only, and as a group they showed more cognitive dysfunction than those without any brain infarcts (75). In a Japanese study of 63 patients with late-onset depression, 59 (94%) had silent cerebral infarcts (76). Finally, in a study of infarcts in the thalamus and basal ganglia, such lesions were found in 14 of 35 depressed elderly patients without neurologic history, but in only 1 of 22 normal elderly volunteers (77).
Depression has been also associated with white matter abnormalities, including white matter hyperintensities (WMHs; areas of increased intensity on T2-weighted MR images) or microstructural abnormalities demonstrated as reduced fractional anisotropy in diffusion tensor imaging. WMHs have been more common in depressed older patients than healthy older adults (10,67,77–81), especially in frontal and temporal regions (82). A post-mortem study has shown that deep WMHs of depressed elderly patients are more likely to be ischemic in nature than deep WMHs of elderly controls (83). WMHs are associated with cerebrovascular disease (84), cardiac disease (85), smoking (86), hypertension (53,84,86), reduced cerebral blood flow (85), executive dysfunction (73,87,88), and disability (52,89).
Based on these findings, the “vascular depression” hypothesis has been proposed, which postulates that cerebrovascular disease predisposes, precipitates, and perpetuates a late-life depression syndrome (3). Vascular disease might lead to depression through damage to specific brain circuits or less directly through inflammation. Proinflammatory cytokines, such as interleukins 1 and 6 (IL1 and IL6) and tissue necrosis factor alpha (TNF-α), are released after damage of the vascular endothelium (90). A post-mortem study found elevated levels of intercellular adhesion molecule-1 (ICAM-1), a marker of ischemia-induced inflammation, in the dorsolateral prefrontal cortex of depressed subjects compared with controls (91). The incidence of depression after treatment with the inflammatory cytokine interferon-alpha (IFN-α) ranges from 30 to 50% (92); and depression has been associated with increases in chemokines, cellular adhesion molecules, acute phase proteins, and proinflammatory cytokines (93,94). Given that proinflammatory cytokines have been associated with atherosclerosis and cardiovascular disease (90), it seems possible that inflammation predisposes to both depression and vascular disease simultaneously; or that inflammation might be involved in a vicious cycle of depression leading to inflammation, which leads to vascular disease, which leads to more inflammation and increased risk for depression.
Beyond inflammation, other factors may set forward a vicious cycle perpetuating depression and worsening vascular disease. These include sedentary lifestyle, overeating, diabetes, smoking, nonadherence with medical recommendations, hypertension, hyperhomocysteinemia, nervous system activation, hypothalamic-pituitary-adrenocortical axis activation and other physiological stress responses, cardiac rhythm disturbances, and hypercoagulability (90,94,95). The relationship between depression and hypercoagulability may in turn be mediated by physical inactivity, smoking, or increased platelet activity (96). Therefore, even in cases where vascular disease predisposes to depression, once depression sets in, a vicious cycle may occur that worsens both vascular disease and depression.
The “vascular depression” hypothesis has served as the conceptual background for further subclassification of geriatric depression. One group of investigators further described a “vascular depression” subtype, subcortical ischemic depression (SID), and defined it as major depression with MRI evidence of subcortical lesions. Unlike most psychiatric disorders, which are described in purely phenomenological terms, subcortical ischemic depression involves a measurable biological abnormality. The association of late-life depression with executive dysfunction led another group of investigators to describe the depression-executive dysfunction syndrome (DED). Although many patients with DED also meet criteria for SID or other “vascular depression syndromes”, DED’s focus on a functional abnormality rather than an anatomical one extends it beyond the vascular depression concept.
Advances in late-life depression research include novel or improved treatments, personalization of treatments according to depression type or characteristics of the individual, and strategies to improve access to and delivery of care. The reader is referred elsewhere for guidelines to treatment of geriatric depression (30,40,97). Here we focus on new developments and trends.
Biological studies of treatment response
Brain research is facilitated by a variety of MRI-based neuroimaging techniques (98,99). Many of these can be performed together in a single MRI scanning session. T1-weighted MRI images allow comparison of brain structure sizes in volumetric brain studies, in addition to classic lesion studies. Brain WMHs can be studied with T2-weighted images, while newer methods allow an examination of white matter tract integrity.
Diffusion tensor imaging (DTI) indicates the direction of water diffusion. Fiber tractography uses DTI information to map out putative white matter fiber tracts. Fractional anisotropy (FA) is a DTI measure of the tendency of water to move in a single direction. Low FA can be a sign of compromised white matter integrity. Another method that reflects white matter integrity, but from a different perspective, is magnetization transfer imaging (MTI), which indicates the amount of water bound to macromolecules such as myelin.
Blood-oxygenation-level-dependent functional magnetic resonance imaging (BOLD fMRI) involves collection of a series of brain images over time, typically 2 seconds apart, so that changes in blood flow reflecting activity throughout the brain can be monitored over time. If a task is performed by the subject in the scanner, usually the goal of the experiment is to identify regions of the brain that are activated in response to the experimental paradigm.
Some techniques allow for analysis of the fMRI data even in the absence of a known experimental paradigm, such as is the case in resting-state experiments, where subjects are not given any task except to “rest” without sleeping. Seed-based methods map out brain regions, whose activity time course is highly correlated with a chosen point or region in the brain. Independent component analysis (ICA) performs a linear decomposition of the fMRI data, treating the data as if it were the sum of numerous components that are spatial maps of brain regions in perfect synchrony with each other. ICA essentially considers the fMRI data to be a symphony of different melodies from different constellations of brain regions and then picks out the melodies being played. In task-driven experiments, ICA has produced results comparable to those derived with a priori knowledge of the experimental time course (100–102).
Some findings employing these techniques have predicted poor treatment response of late-life depression. For example, WMHs in frontal regions were associated with poor response to pharmacotherapy (46) and severity of subcortical gray matter hyperintensities predicted poor response to electroconvulsive therapy (ECT) (45). Lower fractional anisotropy, mainly in frontolimbic areas, predicted poor antidepressant drug response in geriatric depression (103–105).
In adult fMRI studies, lower relative activation of the rostral anterior cingulate cortex (ACC) at baseline in response to negative vs. neutral stimuli was associated with poor response to venlafaxine (106); while decrease in activation (in response to sad faces) of the rostral ACC over the course of treatment with fluoxetine predicted depressive symptom improvement (107).
Novel or improved treatments
ECT remains the most effective treatment for depression. The response and remission rates for untreated late-life depression are up to 90% and 70%, respectively, and for medication-treatment-resistant depression 70% and 50% (108). Some ECT studies have reported higher rates of response and remission for late-life depression than early-life depression (109,110), but this may be due to a tendency to treat elderly depressed patients with ECT before exhausting all other common treatment options.
Much ECT research has focused on minimizing cognitive side effects, such as post-ECT disorientation, anterograde amnesia, and retrograde amnesia. Most cognitive effects are temporary, and scores on cognitive tests generally improve after ECT (due to improvement in depression) (111). However, for some individuals amnesia for events in the days, weeks, months, and in some cases years before ECT does not resolve and can be a distressing side effect.
Varying ECT parameters such as treatment frequency, placement of the electrodes, stimulus energy, and stimulus waveform generally results in a tradeoff between efficacy and cognitive side effects, but some important exceptions to this rule exist. In many studies, greater and faster response at the cost of increased cognitive side effects was obtained by choosing higher treatment frequency, bilateral (BL, bifrontotemporal) rather than right unilateral (RUL) electrode placement, and higher stimulus energy (111).
Some of the best recent research results have been obtained with high-dose RUL ECT. High-dose, brief pulse, RUL ECT resulted in milder cognitive side effects yet equivalent efficacy compared with BL ECT, and greater efficacy compared with lower doses of RUL ECT (112,113). More recently, the effect of ultra-brief pulse width (0.3 ms) combined with either high-dose RUL ECT (6 times seizure threshold) or BL ECT (2.5 time seizure threshold) was compared to traditional brief pulse (1.5 ms) width. The greatest remission rate occurred in patients receiving ultra-brief pulse RUL ECT (73%), and the lowest remission rate in those receiving ultra-brief pulse BL ECT (35%), while traditional pulse width BL ECT (65%) and RUL ECT (59%) had intermediate remission rates.
The ultra-brief RUL group also had less severe cognitive side effects than the other three groups (114). Finally, the efficacy of ultra-brief pulse, high-dose, RUL ECT (6 times seizure threshold) was recently found equivalent to that for ultra-brief pulse, bifrontal ECT (1.5 time seizure threshold), but response for RUL ECT required significantly fewer treatments (7.76 vs. 10.08) (115).
Repetitive transcranial magnetic stimulation
In October 2008, the US Food and Drug Administration (FDA) approved repetitive transcranial magnetic stimulation (rTMS) for treatment of depression resistant to one prior antidepressant medication trial (116). TMS induces flow of electricity at the surface of the brain through powerful, rapidly changing magnetic field pulses. “Repetitive” indicates that more than one pulse is administered. Because the magnetic field strength with conventional coils drops off dramatically with distance from the coil, the effect is much more localized than with ECT.
Unlike ECT, the goal with rTMS is to stimulate without causing a seizure, which eliminates treatment risks involved with seizure and general anesthesia. Neuropsychological and imaging studies have shown that, depending upon the frequency of rTMS pulses applied, rTMS has the capacity to increase (high-frequency pulses, > 1 Hz) or decrease (≤1 Hz) the level of activity in brain circuits at the surface of the brain, which returns to normal within hours (117). For treatment of depression, high-frequency rTMS is usually applied to the (left) dorsolateral prefrontal cortex in an attempt to normalize the low level of activity often found there in depressed patients (118).
Evidence from a recent double-blind, randomized, placebo-controlled trial (N=92) indicates that rTMS may be helpful in the treatment of vascular depression (119). The placebo was a sham coil designed to mimic the skin sensations and noise of rTMS without penetrating into the brain. The rates of response and remission for older patients (age ≥50, mean age 64) with clinically defined (subcortical stroke and/or presence of at least 3 cardiovascular risk factors) medication-treatment-resistant vascular depression were significantly higher for patients treated with a total cumulative dose of 18,000 pulses rTMS (response 39%, remission 27%) than with a sham coil (response 7%, remission 4%).
Magnetic seizure therapy (MST) is the application of high-intensity rTMS to induce a seizure. The main difference between MST and ECT is that MST imparts electrical energy to a limited area on the surface of the brain, while ECT causes electricity to flow throughout the brain as governed by Ohm’s law. This difference might prove advantageous in terms of treatment efficacy or side effects.
Since the first treatment in Bern, Switzerland in May 2000 (120,121), MST’s safety, efficacy, and side effect profile has been explored in small trials. Early MST studies demonstrated a significant antidepressant effect, but this effect was less robust than that of ECT (122), perhaps due to limitations in the maximum energy imparted by the MST device (average 1.3 times seizure threshold) (123). Human testing with a more powerful device began in 2006. With this device, time to recovery of orientation in 11 depressed patients was significantly less (7 minutes after treatment) than for ECT (22 minutes) in the same patients (123).
Vagus nerve stimulation (VNS) typically involves stimulation of the left cervical vagus nerve, with a pacemaker-like device placed in the left chest wall. In animal and human studies, VNS increases cerebrospinal fluid concentrations of neurotransmitters relevant to depression, and alters functional activity of brain regions including orbital frontal cortex, cingulate, thalamus, hypothalamus, insula, and hippocampus (124).
VNS was first used for drug-resistant epilepsy. Studies showed that mood in VNS also improved independently of epilepsy treatment response (125,126), prompting further studies of VNS for treatment-resistant depression. In a 10-week, randomized, controlled trial in drug-resistant depressed patients, the primary outcome measure did not differ significantly between VNS (15%) and sham treatment (10%) (124). Further, it was not clear to what extent the blinding procedure was successful, as VNS even at low intensities causes some subjects to experience physical sensations. However, in the 12-month open continuation phase of the study, where subjects also received treatment as usual (TAU, including medications and/or ECT), treatment response was significantly greater for VNS + TAU (27%) than for TAU alone (13%) (127). VNS for treatment-resistant depression was approved in the European Union in 2001 and by the FDA in 2005 (128).
Deep brain stimulation (DBS) involves electrically stimulating the brain through fine, deeply implanted electrodes. The electrodes typically are attached to a subdermal pacemaker-like device that delivers a continuous train of repeated, very brief, small voltage pulses.
DBS was first used in 1997 for the treatment of Parkinson’s disease (129). Chronic stimulation of portions of the basal ganglia with DBS produced improvements in patients’ motor function similar to those previously achieved with ablation, leading researchers to speculate that DBS produces a reversible “lesion” through transient electrical inactivation (130). Because a decrease in subgenual anterior cingulate cortex (sACC, also called subcallosal cingulate gyrus) activity levels is often associated with depression treatment response, the sACC became the target for the first reported (in 2005) application of DBS for treatment-resistant depression (131). A “striking and sustained remission” was reported in four of six patients, with several patients reporting an almost immediate remission of “painful emptiness” and “void” when the stimulation was turned on. Since then, of 20 depressed patients treated with sACC DBS, response (60%) and remission (35%) rates have been excellent for treatment-resistant (including ECT) depression (132). Post-operative side effects included headache (4 cases), craniotomy site infection (4 cases), and seizure (1 case), but no patient experienced permanent deficits.
The reasons for the good efficacy results are unclear. Some recent mechanistic findings indicate that, although DBS inhibits activity in neuron somata near (<2 mm) the stimulating electrode, DBS stimulates axons, causing activation of efferent nuclei (133). Also, in a more recent study of three persons with extremely resistant forms of depression (one was a 66-year-old woman), DBS to the nucleus accumbens resulted in significant reductions in depressive symptom scores within one week (134). Since increased nucleus accumbens activity has been associated with expectations of and experiences of rewards, it seems unlikely that the improvements in depression were due to inhibition of this nucleus. Results were also encouraging for another study (N=15) involving DBS to the ventral capsule/ventral striatum (VC/VS), with 53% response and 40% remission of treatment-resistant depression. The motivation for applying DBS to the VC/VS was the observation of improved mood in severe obsessive-compulsive disorder patients given this treatment (135).
Forms of psychotherapy reported efficacious in the cognitively intact elderly depressed include interpersonal psychotherapy (ITP) (136), problem-solving therapy (PST) (137), supportive psychotherapy (138), and cognitive-behavioral therapy (CBT) (139,140).
ITP’s focus on loss, grief, and role transitions makes it highly suitable for the elderly depressed population, in whom these themes are common. PST’s structured planning and problem-solving approach should stimulate activity in the dorsolateral prefrontal cortex, a goal of some biological treatments. PST appears effective and well-suited for patients with depression and executive dysfunction (141).
Recently, development of behavioral approaches has begun, based on the concept that depression introduces new problems and at the same time reduces patients’ ability to solve them. As a consequence, depressed older patients experience their environment as difficult to negotiate and stressful, which serves as a trigger perpetuating their symptoms. For this reason, novel treatments have been developed aiming to improve patient adaptation and reduce the experience of adversity. Improving depressed patients’ problem-solving skills is part of several such therapies. However, depressed patients with significant cognitive impairment or physical disabilities may be unable to improve their coping skills with problem-solving techniques. For this reason, we have developed “ecosystem focused therapy” (EFT), a treatment that focuses on the “ecosystem” (patient + environment + family member/caregiver) of which the patient is a part (142). EFT imparts to the patient skills maximizing his/her remaining functions; modifies the patient’s physical environment; and engages family members/caregivers in helping the patient bring to bear his/her skills. EFT uses problem-solving therapy as its framework along with tools and instructions that can be used by patients and caregivers to make the environment conducive to adaptation. Enabling patients to assimilate new skills and changing patients’ environments, including caregivers, to accommodate to the patients’ states, offers patients a good chance at adaptation, increases their sense of mastery, and may reduce depression.
Depression prophylaxis is considered in cases where risk of depression is exceptionally high, such as during the early stages after antidepressant treatment remission or in patients who have had multiple severe episodes of depression (40).
The high rate of depression after stroke has prompted some researchers to explore various forms of post-stroke depression prophylaxis. Prophylactic treatment with escitalopram or problem-solving therapy after stroke resulted in significantly lower rates of depression compared with placebo (143). A large Finnish study compared post-stroke depression rates in districts implementing usual care (mainly physiotherapy and speech therapy) with rates in districts adding an active rehabilitation program. The active rehabilitation group had significantly lower rates of depression (41% vs. 54% at 3 months post-stroke, and 42% vs. 55% after 1 year) (144).
Multidisciplinary approaches to depression treatment
Medical illness increases the risk of depression (95) and depression increases the risk of medical illness. The frequently bidirectional relationship between somatic illness and depression necessitates attention to both somatic and psychiatric issues, in order to break potentially vicious cycles of psychiatric illness predisposing to somatic illness, and vice versa. Non-psychiatric health care practitioners need to incorporate diagnosis and treatment of depression into their routine practice, while psychiatrists to a greater degree need to address somatic issues. Increased collaboration among health care practitioners is likely to improve health care on all levels.
For example, among chronic obstructive pulmonary disease (COPD) patients with depression, both somatic and psychiatric treatment approaches appear to impact both conditions. Depressive symptoms improved after a brief, multidisciplinary inpatient COPD rehabilitation treatment (145). In a study of depressed patients with COPD, physical symptoms, function, and mood improved after treatment with antidepressant medication (146).
Patients with chronic pain and depression are also likely to benefit from a multi-disciplinary approach. Moderate to severe pain is associated with increased rates of depression and poorer depression outcomes; and depression in pain patients is associated with more pain complaints and greater impairment. An estimated 65% of depressed patients have pain and 52% of patients in pain clinics or inpatient pain programs are depressed (147). Underassessment of pain is a major barrier to adequate pain treatment (148); asking depressed patients about pain symptoms and pain treatment can greatly facilitate proper care. Antidepressant medications that increase norepinephrine in synapses, such as tricyclic antidepressants, venlafaxine, and duloxetine, generally help diminish the experience of pain directly, in addition to their antidepressant effects (149).
Finally, the observed bidirectional relationship between depression and vascular disease means that somatic health care practitioners need to assess for depression for the same reasons they routinely assess for hypertension and hypercholesterolemia. Likewise, psychiatrists should consider cardiovascular health in their patient assessments. Such an assessment in psychiatry is particularly important because many psychotropic medications can potentially influence cardiovascular health. For example, by decreasing platelet activity, SSRIs may reduce the risk of heart attacks (150), and also increase bleeding risks in patients taking aspirin.
Psychotropic medications that increase weight, cholesterol levels, or risk of diabetes may also increase risk of atherosclerosis and cardiovascular disease, so many psychiatrists routinely monitor weight and blood pressure; and some prescribe medications such as metformin prophylactically when risk of obesity and diabetes is involved with psychiatric medication treatment (151).
Patients’ concern about weight gain can interfere with treatment, so it is often helpful to choose medications that cause as little weight gain as possible. The medications bupropion (antidepressant) (152) and lamotrigine (mood stabilizer) (153) are generally not associated with weight gain. Among antipsychotic medications, weight tends not to increase much with aripiprazole and ziprasidone (154).
The greatest limitation in treatment of late-life depression concerns treatment access and delivery rather than treatment efficacy. In primary care settings, where most depressed older patients are treated, the diagnosis of depression is often missed. Further, correct diagnosis of depression often does not lead to treatment, and treatment is often inadequate. In one study, only 11 % of depressed high utilizers of primary care treatment were found to receive adequate antidepressant treatment (155).
Studies such as PROSPECT (156,157) and IMPACT (158,159) have shown that collaborative care offered at the primary care setting has superior outcomes to usual care. However, inadequate third-party reimbursements restrict collaborative care to large providers, e.g. health maintenance organizations, which serve a minority of the US population. To overcome this barrier, we proposed a depression care management model (C-DCM) relying on collaboration of primary care physicians with trained social workers employed by community-based, public and nonprofit mental health clinics (142).
While widely available in the US, mental health clinics are rarely connected to primary care practices and underutilized by depressed elders. To utilize this resource, we proposed a collaborative care model, designed to satisfy four conditions. First, it should meet the clinical needs of depressed elders. Second, it should include organizational changes that would enable primary care practices and mental health clinics to work together effectively. Third, collaborative care should be modified in a way that can be used by trained social workers and brings to bear their special skills. Fourth, it should include procedures reimbursed by existing insurance codes so that it adds no cost to primary care physicians or mental health clinics.