- Top of page
OBJECTIVE: Previous treatment trials have found that approximately one third of depressed patients have persistent symptoms. We examined whether depression severity, comorbid psychiatric illness, and personality factors might play a role in this lack of response.
DESIGN: Randomized trial of a stepped collaborative care intervention versus usual care.
SETTING: HMO in Seattle, Wash.
PATIENTS: Patients with major depression were stratified into severe (N = 149) and mild to moderate depression (N = 79) groups prior to randomization.
INTERVENTIONS: A multifaceted intervention targeting patient, physician, and process of care, using collaborative management by a psychiatrist and primary care physician.
MEASUREMENTS AND MAIN RESULTS: Patients with more severe depression had a higher risk for panic disorder (odds ratio [OR], 5.8), loneliness (OR, 2.6), and childhood emotional abuse (OR, 2.1). Among those with less severe depression, intervention patients showed significantly improved depression outcomes over time compared with those in usual care (z = −3.06, P < .002); however, this difference was not present in the more severely depressed groups (z = 0.61, NS). Although the group with severe depression showed differences between the intervention and control groups from baseline to 3 months that were similar to the group with less severe depression (during the acute phase of the intervention), these differences disappeared by 6 months.
CONCLUSIONS: Initial depression severity, comorbid panic disorder, and other psychosocial vulnerabilities were associated with a decreased response to the collaborative care intervention. Although the intervention was appropriate for patients with moderate depression, individuals with higher levels of depression may require a longer continuation phase of therapy in order to achieve optimal depression outcomes.
The development of the Agency for Health Care Policy (AHCPR) Guidelines for Treatment of Depression1 has stimulated researchers to test new intervention models that attempt to improve the process and outcomes of care for primary care patients with major depression. Controlled, randomized trials of implementation of the ACHPR guidelines for antidepressant therapy have demonstrated that the guideline recommendations, which were based largely on specialty clinic efficacy studies, can be extended into primary care settings.2–4
One of the important differences between primary care and specialty clinic settings is the wider range in severity of depression present in patients in primary care.5 More patients with milder forms of major depression are likely to be enrolled in primary care studies, and these patients are likely to have less psychiatric comorbidity.5,6 The wider range of depression severity in primary care settings may increase the potential variability of outcomes. Important differences in outcomes between severe and less severe forms of illness may be obscured unless this variability is taken into account. Even in psychiatric populations, there appear to be differences in outcome between treatment trials with patients having mild and severe forms of major depression. This was illustrated in the National Collaborative Study of Depression where patients with milder forms of major depression responded to both active and placebo treatments, while those with more severe forms of illness had significantly greater clinical improvement with medications or psychotherapy.7
Most antidepressant or psychotherapy intervention trials have found that approximately one third of depressed patients have persistent symptoms or treatment resistance.8,9 Little is known about the predictors of this nonresponse in primary care settings. In a previous primary care study, we found that the best predictors of negative outcomes for patients with depression were the initial severity of depression based on the 20 depression items of the Hopkins Symptom Checklist and higher neuroticism scores.3,4 Clinical studies of patients with major depression have also shown that depression severity and neuroticism may also be associated with additional comorbidity such as panic disorder and adverse experiences in childhood.10
Using an intent-to-treat analysis, we recently reported that our collaborative care intervention resulted in improved depressive outcomes, compared with usual primary care in patients with persistent symptoms approximately 2 months after initiating treatment with their family doctor.11 Because of prior research describing variability in outcome by severity of depression,3,12 we modified the current primary care clinical trial by using a stratified randomization scheme that balanced severity of SCL-20 ([Hopkins] Symptom Checklist-20) depression between intervention and control groups, using a mean item score of 2.0 as the cut point. We hypothesized that compared with patients with higher depression scores, those with lower depression scores would have significantly greater intervention versus control differences in depression outcomes. We also collected baseline information about early family environment, current stressors, family history of psychiatric illness, the personality trait of neuroticism, and social support. We further hypothesized that early adverse experiences (e.g., childhood maltreatment) as well as sequelae frequently associated with these traumas (e.g., anxiety, loneliness, and high neuroticism) would be associated with higher SCL scores.
- Top of page
As shown in Table 1, patients with mean item SCL scores greater than 2.0 were significantly younger (t 226 = 2.46, P = .02), had higher neuroticism scores (t 225 = 2.09, P = .04), had the onset of their first depressive episode at an earlier age (t 226 = 3.47, P < .001), and were significantly more likely to have comorbid panic disorder (χ21 = 13.79, P < .001) compared with patients with SCL scores from 1.0 to 2.0. As would be expected, patients in the two SCL strata had significantly different baseline SCL scores (t 226 = 19.31, P < .001). The SCL severity groups did not differ in gender, education, employment, race, chronic disease score, or percent of patients with recurrent depression or dysthymia.
Table 1. Demographic and Clinical Characteristics of Patients in the Depression Strata
|Variable||Low SCL (n = 149)||High SCL (n = 79)|
|Mean age, y*±SD||48.6 ± 14.1||43.9 ± 12.4|
| ≥1 year of college, %||78.5||75.9|
|Employed full- or part-time, %||65.5||74.7|
|Chronic Disease Score, mean ±SD||1380.3 ± 1154.4||1090.4 ± 1116.9|
|NEO Neuroticism, mean ±SD*||22.3 ± 5.3||23.9 ± 5.8|
|SCL-Depression, mean ±SD†||1.6 ± 0.3||2.5 ± 0.4|
| % with recurrent depression (3 or more episodes)||78.9||84.7|
| % with dysthymia||46.9||41.8|
| % with panic disorder‡||4.0||20.3|
| % of first-degree relatives with depression, mean ±SD||41.8 ± 28.5||36.6 ± 26.6|
|Age of onset of first depression, y†±SD||38.0 ± 17.9||29.7 ± 15.8|
Patients with SCL scores greater than 2.0 rated themselves as having significantly more stress over the past 3 months (t 225 = 2.44, P = .02) and stated that the stress interfered with their daily activities more than patients with less-depressive severity (t 226 = 2.86, P = .004) (Table 2). They also rated themselves as significantly more lonely (t 226 = 4.03, P < .001) and as having more childhood emotional abuse (t 226 = 4.02, P < .001). The groups did not differ on any other social support or vulnerability variables. There were no differences between the groups with respect to the percentage of patients seeing a mental health provider, number of primary care visits for depression, number of visits for other medical problems, or number of visits or telephone calls related to the intervention over the 6-month time period.
To summarize the univariate findings, age, neuroticism, stress level and interference from stress, comorbid panic disorder, age of onset of first depression, loneliness, and childhood emotional abuse were allowed to enter in a stepwise fashion into a logistic regression model predicting SCL depression group severity strata. Both backwards and forwards methods were employed. Both techniques produced the same model with three significant independent predictors of SCL severity status: comorbid panic disorder (Wald's t = 11.14, P < .001), loneliness (Wald's t = 9.46, P < .002), and childhood emotional abuse (Wald's t = 5.42, P < .02). Patients in the high severity SCL group were more likely to have comorbid panic disorder (odds ratio [OR], 5.8), to feel more lonely (OR, 2.6), and to have experienced childhood emotional abuse at least “sometimes” (OR, 2.1), compared with patients in the lower SCL group.
As seen in Figure 1, intervention patients in the lower SCL severity group were significantly more likely to improve over time compared with usual care patients (z = −3.06, P < .002). Conversely, Figure 2 shows that in the more severely ill depression group, there was no significant group-by-time effect (z = 0.61, P = NS), although both intervention and usual care groups had a significant decrease in depressive symptoms over time (z = −7.09, P < .001). Interestingly, in the more severe group, collaborative care appears to produce a more significant change in first 3 months (similar to the differences found between the less severe intervention vs control groups), yet that difference disappears between 3 and 6 months.
Figure 1. SCL depression adjusted means (adjusted for age, gender, chronic disease score, and neuroticism) with standard errors for patients in the high severity strata.
Download figure to PowerPoint
Figure 2. SCL depression adjusted means (adjusted for age, gender, chronic disease score, and neuroticism) with standard errors for patients in the low severity strata.
Download figure to PowerPoint
Table 3 presents the percentages of adequacy of antidepressant dose, adherence to medications, and the percentage of patients taking selective serotonin reuptake inhibitors (SSRIs) in the care as usual and the intervention groups classified by depression severity. Adequacy is shown in both the lowest dose, consistent with ACHPR guidelines, and twice the lowest dose. Intervention patients in both severity groups received more intensive pharmacotherapy and were significantly more likely to receive an adequate dose of antidepressants for 90 days or more compared with usual care controls. Despite the lack of significant effects on depressive symptoms between intervention and usual care patients in the high severity group, self-report data show that the intervention patients in the high severity group were significantly more likely to have taken medication for 25 or more days during the prior month at the 6-month follow-up. Compared with the usual care patients, the intervention patients in the high severity group were also more likely to be taking SSRIs at the 6-month follow-up.
Table 3. Adequacy and Adherence Percentages for Intervention and Control Groups Within the Low and High Depression Strata
| ||Low SCL Group (n = 149)||High SCL Group (n = 79)|| χ 2 (df = 1)|
|Controls, %||Intervention, %||Controls, %||Intervention, %||Low||High|
|Patients Taking SSRIs|
To investigate whether differences in antidepressant adherence between intervention and usual care patients might explain the disappearance of treatment response in the high depression severity intervention group (Fig. 2), we examined automated data on refills during this period. We compared intervention and controls in the high severity group with respect to gaps of 15 days or more in their antidepressant adherence for the interval between 3 months (the end of the intervention) and 6 months. We found that, compared with usual care patients, significantly fewer intervention patients in the high depression severity group had refill gaps of 15 days or more (38.7% vs 61.3%). This suggests that the intervention patients had a more continuous adherence to medications than controls between 3 and 6 months, despite having increased depression symptoms.
- Top of page
In a previous paper describing the main results of this randomized trial of collaborative care intervention for primary care patients with depression, we reported that the intervention was significantly more effective than usual care in adherence to guideline-level antidepressant treatment (25 days or more out of each month at each follow-up), satisfaction with care of depression, antidepressant dosing adequacy, and depression outcomes.12 The present analysis took the next step and asked whether depression severity and psychosocial variables related to severity were factors associated with response to the intervention. By using the balanced randomization categories of the original study (low and high severity depression) as a classification variable, we demonstrated that the main effect of this intervention occurred in patients with lower severity depressive symptoms (Fig. 1), with no significant intervention versus usual care differences over time for patients with more severe forms of depression at baseline (Fig. 2).
Although it is not surprising that individuals with more severe affective illness might have a less robust response to the intervention, delineation of the specific factors associated with this lack of response is important in that it may lead to a better understanding of why some patients fail to improve and how to direct additional resources to their care. We found that patients with increased depression severity demonstrated not only less response to treatment but also higher risk for current comorbid panic disorder and psychosocial factors, such as decreased social support and childhood emotional abuse. These findings suggest a clinical model in which interactions with the intervention provider and the effects of preexisting vulnerabilities might account for the decreased response in these more severely ill individuals and infer treatment approaches that could boost the intervention effect in selected patients.
The first part of the model hypothesizes that the study psychiatrists may be partially responsible for the changes shown in Figure 2. It appears that in patients with more severe depression, there is a strong trend for intervention patients to improve faster than controls between baseline and 3 months (the period during which the intervention was occurring), but between 3 and 6 months (after cessation of the psychiatrist visits), the intervention patients seem to develop worsening depressive symptoms.
The disappearance of this effect from 3 to 6 months was studied in further analyses. One possibility was that the intervention group may have decreased adherence to antidepressant medication after the psychiatry visits ended (in most cases by 3 months). However, analysis of refill patterns from automated data showed no differences in adherence (as measured by refills) between the high and low depression severity groups for the period of 3 to 6 months, suggesting that the worsening depressive symptom course in intervention patients was not due to medication adherence.
The second possibility, more consistent with the data, is that more severely depressed patients may need longer follow-up periods of specialty mental health care. In the high depression group, there was a significant improvement in intervention versus usual care patients in adequacy of pharmacotherapy for 90 days or more at both the lowest range of antidepressant medication dosage and twice the lowest range, as well as marked differences in adherence between interventions and controls at the 6 month follow-up (75% vs 39%). Intervention patients were also significantly more likely to receive SSRI prescriptions compared with usual care patients at 6 months. These differences in adequacy of pharmacotherapy coupled with the pattern of change in Figure 2 suggest that it may be the combination of the support of the psychiatrist and antidepressant medication that may be associated with the initial change in depressive symptoms between baseline and 3 months; however, once the support of the psychiatrist was discontinued, improved pharmacotherapy alone was inadequate in maintaining improved outcomes. In contrast, patients with less severe depression (Fig. 1) seem to be able to maintain their initial significant gains with primary care visits and antidepressants alone.
The influence of psychiatric comorbidity and preexisting vulnerability factors is also consistent with this explanation. Patients with early childhood maltreatment are at higher risk for later psychological distress, especially anxiety, depression, low self-esteem, problematic adult social relationships, and impaired social skills.25 The results of the logistic regression showing that the high severity group was significantly more lonely and had a higher likelihood of comorbid panic suggest that the combination of anxiety and loneliness may have made these patients more vulnerable to relapse after stopping the additional visits with the psychiatrist.
One of the most robust clinical predictors of differences between the two groups of patients was the presence of comorbid panic disorder. Panic disorder was approximately 5 times as prevalent in patients in the more severe depression subgroup. Several research groups have found that primary care patients with comorbid panic disorder have increased severity of depressive symptoms. For instance, Lecrubier et al.26 found that primary care patients with major depression and panic disorder compared to those with major depression alone had higher General Health Questionnaire (GHQ)-28 scores (17 compared with a score of 12 for depression alone). Moreover, those with comorbid panic and/or agoraphobia and major depression had significantly more disability days per month (12 vs 7.2) than those with depression alone, had an earlier age of onset, and were more likely to have long-term disorders. Several other primary care researchers have reported that patients with comorbid panic disorder and major depression have greater physical and psychosocial functional impairment compared to those with major depression alone.27–31 Brown and colleagues have shown that primary care patients with comorbid lifetime panic disorder and major depression were more likely to have a history of alcohol dependence, somatization disorders, and avoidant personality disorder.27
Brown and colleagues30 have also shown that the presence of comorbid panic disorder in primary care patients with major depression was a significant predictor of poor outcomes in a trial of antidepressant medication versus interpersonal therapy versus usual primary care. Depressed primary care patients with lifetime panic disorder were significantly more likely to prematurely terminate pharmacotherapy or psychotherapy during each treatment's acute phase. The presence of comorbid panic disorder in patients with major depression has also been shown to predict negative outcomes of treatment with antidepressant medication in efficacy trials,31,32 and has been associated with early childhood abuse and neglect.10
There are several possible limitations of this research. We tested patients with high rates of employment, with 1 or more years of college, and who were members of an organized system of health care with limited racial and ethnic diversity. However, effectiveness trials have shown that this collaborative model of care can be successfully adapted to patients with a variety of socioeconomic backgrounds and diverse practice settings as well.33,34 Because referral to the study was dependent upon the accuracy of primary care physician diagnoses, the study may not be generalizable to the 40% to 50% of patients who do not receive accurate diagnosis in primary care or the patients identified by the physician as depressed but not receiving medication. In addition, the measurement of psychosocial vulnerability was a secondary goal of this study, and the findings of emotional abuse and social isolation as important contributors to poor response should be confirmed in larger well-designed studies using the full scales from which these items were derived.
Our findings suggest that the collaborative care intervention improves the care and long-term depressive outcomes for patients with persistent symptoms of moderate severity after primary care treatment. The more severely ill patients may require more intensive clinician follow-up and/or psychotherapy to achieve sustained improvement from depressive symptoms. The widely held clinical impression that patients with more severe depression may need more intensive clinical management was recently examined in a “mega-analysis” of depression treatment studies9 where combined antidepressant and psychotherapeutic approaches were found to be superior to either therapy alone for patients with severe, recurrent depression. These findings suggest that the testing of stepped care algorithms that move more severely depressed patients into increasingly intensive treatment protocols is the next step in the cost-effective treatment of patients with recurrent, severe major depression.