While a great deal of attention has been devoted to racial and ethnic health disparities, health disparities in individuals disadvantaged by low income, lack of education, or employment (i.e., lower class) have been relatively neglected (1). Depression and anxiety disorders are known to be more prevalent in economically disadvantaged populations (2–5), and impoverished depressed and anxious persons have demonstrated poor clinical outcomes in observational studies (2, 4). Despite these associations, little is known about outcomes of evidence-based care with poor and low-income individuals. Miranda and colleagues (6) showed that both cognitive behavior therapy (CBT) and antidepressant medications (paroxetine or bupropion) were effective for treating poor minority women with depression who received extensive outreach and supportive services for attending care. A study of impoverished Chilean primary care patients with depression demonstrated that a collaborative care program designed to increase adherence to antidepressant pharmacotherapy improved outcomes (7). While both studies document the effectiveness of evidence-based interventions relative to usual care in poor populations, neither study included a comparison group of subjects with greater socioeconomic resources. In a secondary analysis of a large (43 primary care clinics) quality improvement study for depression (Partners in Care), the quality improvement interventions were found to produce significantly greater effects in Latino and African American patients than in Caucasian patients (8). Although the Latino patients tended to be poorer than the Caucasian patients, the impact of poverty per se on outcome was not examined.

We previously conducted a three-site primary care study (the Collaborative Care for Anxiety and Panic study) that demonstrated superior clinical effectiveness of a combined CBT and pharmacotherapy intervention relative to usual care for panic disorder patients in six university-based primary care clinics in three West Coast cities (9, 10). Because approximately one-third of this study group was at or below the federal poverty line and included significant numbers of impoverished Caucasian patients, this study provides an opportunity to examine the association between income and clinical outcomes without substantial confounding by ethnicity. Furthermore, because disadvantaged patients are more likely to seek care in medical as opposed to mental health specialty settings (11), this study group and setting is especially appropriate for examining the effects of poverty. The purpose of this secondary post hoc analysis is to determine whether or not, and to what extent, economic disadvantage is associated with clinical outcomes when an evidence-based intervention for panic disorder is compared with treatment as usual. If poor subjects show less improvement than those who are not disadvantaged, it is likely that interventions tailored to the needs of this population will be required to optimize the effectiveness of these interventions. If poor subjects demonstrate a better or equal response, then specially tailored interventions are unlikely to be required.

Setting and Subjects

Patients between the ages of 18 and 70 who met DSM-IV criteria for panic disorder—with no comorbid psychiatric or medical conditions that were life threatening (i.e., suicidal ideation, terminal medical illness) or likely to interfere with participation (psychosis, dementia, severe substance abuse)—were recruited from six university-affiliated primary care clinics in Seattle, San Diego, and Los Angeles. Consistent with effectiveness design (12), all subjects meeting inclusion criteria were enrolled regardless of prior treatment history or current use of medication or psychotherapy. The study was approved by the institutional review boards of all three universities (University of Washington, UCLA, and UC San Diego).

Procedure

More detailed descriptions of the assessments and interventions for the Collaborative Care for Anxiety and Panic study have been previously published (9, 10, 13). After baseline assessment, subjects were randomly assigned to the intervention or usual care condition. For the intervention, a master’s-level behavioral health specialist delivered, free of charge, six sessions of CBT for panic disorder and related depression and anxiety symptoms (14, 15) and also coordinated the patient’s care, which included medication managed by the primary care physician using an algorithm (16) and paid for by the patient. Specific medication recommendations for individual subjects were relayed as needed from a consulting psychiatrist to the primary care physician via the behavioral health specialist, who informed the psychiatrist about subject clinical status and received feedback at weekly meetings. Subjects had to complete the six CBT sessions within the first 3 months of the study. Six follow-up telephone booster sessions, each lasting from 15 to 30 minutes, were scheduled through the rest of the year at 6- to 12-week intervals to monitor clinical status, reinforce proper medication use and cognitive behavior skills, and make further medication recommendations if necessary. Subjects in the usual care condition received pharmacotherapy from their primary care physician without psychiatric consultation, with referral to specialty mental health providers possible. Hence, in this effectiveness study, treatments were not matched in the two study arms but allowed to vary according to subject adherence and choice. Patients assigned to the usual care condition could conceivably receive CBT from specialty referral sources.

Assessments were conducted by telephone with interviewers blind to subject intervention status administering questionnaires at baseline and every 3 months during the course of the study. This secondary analysis focused on dimensional measures of panic/anxiety, depression, and functional disability as determined with 1) the Anxiety Sensitivity Index (17), a scale that measures the cognitions that underlie panic-related somatization but is also sensitive to the frequency of recent panic attacks (18); 2) the CES-D (19), a scale measuring depression severity (19); and 3) five items selected from the larger WHO Disability Scale (20). The Anxiety Sensitivity Index was chosen as a simple global measure of panic on the basis of previous consensus that panic frequency as a single measure of panic correlated poorly with global burden of illness and outcome (21). We also measured intervention effects on subject-reported quality of pharmacotherapy (guideline-concordant antipanic medication at a sufficient dose for at least 6 weeks [16]) and psychotherapy (attending a minimum of three sessions that contained a minimum of four of seven key components considered characteristic of CBT [22]). Our measure of income was derived from self-report of annual income and family size at baseline telephone assessment.

Statistical Analysis

The overall study group was divided into those patients with income at or below the poverty line (N=80) and those with income above the poverty line (N=152). We used 2001 Federal Poverty Guidelines (23) and indexed the categorization to family size reported by subjects. Baseline demographic and clinical characteristics of patients above and below the poverty line were compared using group t tests. For the longitudinal analysis of intervention effects, we specified a (2-level) hierarchical model with random effects, using all patients randomly assigned to a condition. The repeated observations model (or level 1 model) was a piecewise linear growth model (24) that specified a linear segment between baseline and the first follow-up evaluation at 3 months (at which point the six CBT sessions ended) and then another linear segment for the subsequent follow-up evaluations at 6, 9, and 12 months. This model was intended to reflect an observation from previous effectiveness studies in which the greatest effect occurs during the acute study phase, mirroring the greater intensity of early intervention, and then effects remain stable, fall off, or increase at a much diminished rate. Time trends for patients within group were allowed to vary around the group-specific mean by the inclusion of patient-specific random effects for the intercept (i.e., baseline), the first slope (i.e., baseline to 3 months), and the difference between the second slope (i.e., 3 months to 12 months) and the first slope. We adopted a Bayesian approach (25, 26) to fit this model to both continuous and dichotomous longitudinal responses using baseline, 3, 6, 9, and 12-month follow-up data. We used previously described methods (27, 28) and WinBUGS software (29) to implement the model. The model provided 95% posterior probability intervals for main effects of intervention, income, and their interaction.

There were significant and substantial differences between subjects above (N=152) and below (N=80) the poverty line at baseline time of study entry (Table 1). Impoverished subjects were somewhat older, had less education, more self-reported chronic medical illnesses, more frequent full-symptom panic attacks, higher Fear Questionnaire scores, greater anticipatory anxiety and anxiety sensitivity, higher CES–D scores, higher rates of current PTSD and comorbid agoraphobia, poorer physical functioning and emotional well-being, and a higher rate of receiving any mental health specialty care in the preceding 3 months. However, differences in ethnicity (fewer impoverished patients were Caucasian) and receipt of recent treatment with medication or psychotherapy (somewhat higher in impoverished patients) were not statistically significant.

Effects of Intervention on Guideline-Concordant Care

We previously reported that for the entire population included in this study, our intervention did not result in different rates of guideline-concordant pharmacotherapy relative to usual care but did result in increased rates of CBT treatment at the 3-month time point relative to usual care (10). The results of new analyses that focused on the slope or rate of change between baseline and 3 months and between 3 months and 12 months demonstrate again an intervention effect for CBT (between 0 and 3 months) but not medication (Table 2). More important, however, the analysis shows that there is no income effect and no interaction between income and the intervention. That is, the effects or lack of effects of the intervention on rates of guideline-concordant care hold true for subjects both above and below the poverty line. Thus, we can conclude that income is not confounded in this study with the receipt of care that was more likely to be guideline-concordant or more clinically appropriate.

Outcome Effects

Effects on the various outcome measures are included in Table 2 and Figure 1, Figure 2, and Figure 3. Table 2 again depicts regression parameters for the baseline, 0–3-month, and 3–12-month slopes for each outcome. As documented in our previous study (10), there were significant intervention effects on dimensional measures of anxiety sensitivity, depression severity, and disability. Consistent with the baseline data, the two income groups significantly differed on all three measures at baseline (intercept), with impoverished patients doing more poorly. Overall, impoverished patients, despite showing greater improvements in intervention compared with usual care, continued to show greater symptom severity over the yearlong study than patients above the poverty line. Significant differences were observed between the two income groups for all three measures for the acute time trend segment (baseline to 3 months) and significant differences for anxiety sensitivity and depression severity but not disability for the more extended time trend segment (3–12 months). There was no differential effect of the intervention in the two income groups (i.e., no income-intervention interaction) at either of these two time-trend segments (i.e., the acute 3 month phase or the longer 3–12 month phase).

This analysis confirms prior findings that low-income psychiatric patients with depression and anxiety have more severe symptoms and functional disability associated with their illness (2, 3, 5, 30). Even after they receive clinically effective, carefully executed, evidence-based interventions their outcomes, because they start out more severely ill, lag behind those without economic disadvantage. At baseline, our panic disorder patients with incomes below the poverty line had more severe symptoms of panic, anxiety, and depression and also more severe chronic medical illness and associated physical disability. This may be related to the increased rate of stressful life events (31, 32) and exposure to trauma (33) in the poor, along with a relative lack of adequate resources to cope with these stressors (34, 35). This is also consistent with a previous study contrasting primary care panic patients from a public sector setting with panic disorder patients from more typical primary care settings (30). In that study, the low income and minority status of patients seeking care in the public sector setting accounted for the majority of clinical and functional differences (i.e., these differences disappeared when these variables were included as covariates) except for anxiety symptoms, which may have been affected by additional unmeasured factors related to the disadvantaged states of patients served in this setting (e.g., the chronic stress associated with dangerous neighborhoods and diminished social and material resources).

There were no differences in the effect of our intervention on quality of care in the two income groups. In both groups, the intervention successfully improved exposure to quality CBT during the first 3 months, when CBT sessions were offered to intervention patients. In contrast, the intervention failed to improve the proportion of patients receiving high-quality pharmacotherapy, with both intervention and usual care patients in both income groups increasing their exposure to quality pharmacotherapy during the study. It is possible that the nonmedical background of the behavioral health specialist—or the competing demands of both delivering CBT and trying to maximize medication use—may have led to less than optimal focus on or achievement of quality medication. It is also possible that patients were less motivated to pay for and maximize their use of medications when CBT was already improving their symptoms and was being provided free of charge.

Finally, the intervention had equivalent effects on the various measures of clinical and functional outcome in the two income groups. Low-income patients started off with more severe symptoms and on average remained more symptomatic than patients above the poverty line, with both income groups showing greater improvement with the intervention relative to usual care. It is of interest that low income patients receiving the intervention were able to achieve a level of anxiety symptom reduction on the Anxiety Sensitivity Index comparable to higher income patients in the usual care group, suggesting that this intervention was able to eradicate the differences in illness severity that existed at baseline and also persisted across the combined group (intervention plus usual care) at 12 months.

Previous analysis from this dataset documented that more patients in the intervention group received CBT than in the usual care group, but that delivery of high-quality pharmacotherapy was not different in the two groups (10). CBT appeared to be the active ingredient in the intervention. The current analysis, demonstrating similar intervention response in subjects above and below the poverty line, suggests that CBT as an intervention for panic disorder works equally well in poor people. Thus, we do not think that this intervention needs to be modified or specially “tailored” for this population, even if these individuals may be less well educated. This is consistent with recent studies showing that CBT can be effectively delivered to anxiety disorder patients in low income settings (36) and to low income panic disorder patients of African American ethnicity (37), although none of these studies included an inactive treatment control group so that magnitude of treatment response could be compared. It is also consistent with studies documenting an expected level of response to evidence-based mental health treatments in disadvantaged depressed patients (6, 7). Unlike the Partners in Care study, where depressed minority patients (many of them quite poor) improved more than depressed Caucasian patients (most of whom were not poor), our poor patients appeared to respond equally to our more economically advantaged patients. This may be because, unlike that study, our poor patients’ rate of receiving care at baseline was comparable to, rather than lower than, the comparison group. It should be noted, however, that our power to detect an income-intervention interaction, which is estimated in this study to be quite small, is extremely low and would require a much larger sample size. It would not appear, however, that this estimated small effect would likely be clinically significant.

In conclusion, this analysis showed that impoverished patients with panic disorder responded equally well to an intervention designed to increase rates of quality antipanic pharmacotherapy and CBT. Because these patients began with more severe symptoms and disability, they ended up after treatment with less complete responses (i.e., a higher level of residual symptoms) than patients above the poverty line who started off with less severe symptoms. Their greater rate of comorbid psychiatric and medical conditions also likely contributed to this. This suggests that although CBT worked equally well in this impoverished population, helping these initially more ill patients achieve more complete resolution of their symptoms and functional deficits will require more intensive treatment, characterized either by greater amounts and frequency of the same treatment or perhaps additional “stepped-care,” which would deliver either adjunctive medications or behavioral treatments targeted toward both residual symptoms and additional comorbid conditions that may be contributing to disability. This hypothesis should be tested in future controlled studies.