The BTB study protocol received ethical approval from the Institutional Review Board at Thomas Jefferson University (Control #06F.551) in which the protocol was initiated; with ethical approval also subsequently received by the Institutional Review Board at Johns Hopkins University (Control # NA-00046775) upon appointment of the Principal Investigator and lead author. Also, a Federal Wide Assurance Agreement between the senior center agency (Center in the Park) in which the study was implemented and Thomas Jefferson University was obtained.
BTB reflects a partnership between a university research center and a senior center with each contributing to the plan and execution of the study and model design while preserving internal validity and control of a randomized trial. Agency staff are trained in human subjects and participate in recruitment efforts, conduct depression screenings, and deliver the home intervention; research staff provide oversight of trial procedures, jointly participate in recruitment efforts, train agency staff in depression screening, implement randomization procedures, and provide oversight of intervention delivery and monitoring of treatment fidelity. The design optimizes experimental control while building capacity of the participating senior center to screen and deliver BTB; hence, maximizing potential for sustainability and self-sufficiency at trial conclusion if BTB is efficacious and cost-effective. Additionally, the partnership enables the senior center to access mental health researchers and experts to provide oversight and guide implementation. The academic institution maximizes recruitment and trial success through involvement upfront of end users of the intervention. This approach may also enhance sustainability if the intervention is shown to be effective.
General procedures include two sequential depression screenings of individuals by senior center staff over two weeks using the Patient Health Questionnaire (PHQ-9). Those eligible (PHQ-9 ≥ 5) and willing provide written consent using an approved Institutional Review Board (IRB) form (Thomas Jefferson University IRB control #06f.551) and receive a baseline home interview conducted by academic partners. Participants are then randomized to receive BTB immediately or 4 months later (wait-list control). All participants are reassessed at 4 and 8 months by academic partners who remain masked to study assignment. The design allows for a true two-group comparison at 4 months to evaluate intervention effects on symptom severity (primary end point), quality of life indicators, anxiety, depression knowledge, and functional disability (secondary endpoints); an evaluation of long-term (8 months) effects for the initial treatment group; and whether wait-list controls derive similar benefits from 4 to 8 months. Treatment fidelity is enhanced through use of treatment manuals, session implementation checklists, audiotaping and rating of randomly selected sessions, one-on-one supervision and bi-weekly group debriefing sessions involving structured case presentations and review of treatment documentation.
Aims and study hypotheses
Our primary aim is to test the immediate effect of the intervention at 4-months on symptom severity (PHQ-9 severity score) in African American older adults (primary trial outcome; between group comparison). Our hypothesis is that participants in the initial treatment group will report fewer depressive symptoms in comparison to control group participants receiving usual care. We also seek to test the maintenance effect of the intervention at 8-months on depression (within group comparison). Our hypothesis is that participants in the initial treatment group will maintain reduced symptom presentation from 4 to 8 months and the wait-list control group will show similar reductions in depressive symptoms as the initial treatment group (within group comparison).
A secondary aim is to evaluate whether BTB has immediate (4 months) and long-term (8 months) effects on quality of life indicators, functional difficulty, anxiety, and depression education and behavioral activation levels. We expect that the initial treatment group will show improvements in these areas relative to the wait-list control group, that improvements will be maintained at 8 months and that the wait-list control group will demonstrate similar improvements at 8 months following their receipt of the intervention.
Several exploratory aims are also proposed. We will evaluate the mechanism of action, or pathways, by which treatment gains are obtained. Given that behavioral activation represents conceptually the key active ingredient of the proposed intervention, we plan to evaluate its mediational effect. We will also evaluate whether the intervention has a differential treatment effect based on a study participant's gender, age, and living arrangement (alone or with others). Given that previous research suggests that participant characteristics may moderate depressive symptoms and treatment outcomes, these exploratory analyses will provide insight as to whether this particular treatment benefits some groups more than others.
Recruitment, eligibility and randomization
Participants enrolled in the trial were recruited from two sources, a program providing short-term in-home support for individuals who are medically comprised (referred to as the In-Home Support Program; IHSP) or the community at-large. IHSP provides homebound elders assistance regaining independence after illness or hospitalization through home health and homemaker services, home delivered meals, transportation, or companions but does not typically involve formal and systematic depression screening. Recruitment from IHSP involved integrating depression screening in the assessment process of participants in the program. Recruitment to the community at-large involved print media, presentations to community groups, and print media.
Eligibility criteria for BTB include being African American; ≥ 55 years of age; English speaking; cognitively intact (MMSE > 24), and a score ≥ 5 on the PHQ-9 on two sequential testing occasions. Other criteria are practical and include having a home telephone and planning to live in the area for 8 months. Individuals are not eligible with serious mental illness histories, life-limiting illnesses, involvement in another clinical depression trial, or if living in assisted living or nursing home facilities. Individuals receiving anti-depressant medications or other mental health treatments are eligible.
As individuals from IHSP may be more medically and functionally compromised than those enrolled from the community at-large, we used a stratified randomized design to assure equivalence between groups along this dimension. Permuted block randomization with each stratum (IHSP versus other recruitment source) to control for possible changes over time in subject mix was employed. The blocking number was developed by the trial senior statistician and was not disclosed to the investigative team.
In selecting measures for this study, we chose those with known reliability and validity; sensitivity to change that is expected to occur as a consequence of intervention; relevance to and wide usage in depression research, and acceptability by the target population. Our primary outcome measure is the Patient Health Questionnaire (PHQ-9), a brief, 9-item self-report measure specifically developed as an assessment tool for the diagnosis of depression in primary care. In addition to its psychometric soundness and demonstrated reliability, validity, sensitivity and specificity, the measure is also widely used in community-based settings and research trials. It yields a depression severity score (0 to 27) and a diagnostic category (minimal to none, mild, moderate, moderate severe, severe) mapping on to the 9 DSM-IV. Administration involves initially asking two items (loss of interest; feeling upset, distressed or depressed). Individuals reporting that either symptom occurs more than half the time over past two weeks are administered 7 other items [29
]. Thus, two scores can be derived: symptom severity and categorical diagnosis. We also consider as a complementary secondary outcome a change in categorical diagnoses in which we seek to determine the number of persons who shift to none to minimal diagnostic category at 4 months. A categorical approach provides a clinical interpretation of treatment efficacy and also whether the intervention addresses certain syndromes better than others (e.g., minor depressive disorder may be more responsive to the proposed intervention than major severe depression). While the PHQ-9 is a self-report measure, self-report has been shown to be highly associated with clinical determinations. Also, the measure has been shown to be effective when administered in person or via telephone. The secondary outcome measures for the trial are presented in Table .
Secondary Outcome Measures for BTB Trial
Up to 10 one hour sessions occur over 4 months. A home visit can be replaced with check-in telephone calls once steady progress is demonstrated. Sessions integrate five treatment components previously shown to be effective and modified in BTB to resonate with cultural preferences: case management, referral and linkage, depression education, stress reduction techniques, and behavioral activation. (Table ) Critical is tailoring to participants' depression knowledge level, care management needs, and self-identified goals and behavioral activation plans.
Intervention Component, Content, and Modification for Target Population
Initial sessions occur weekly and then bi-weekly. In session one, building rapport and a therapeutic relationship begins; also, interventionists assess care management needs and begin education about depressive symptoms making connections between behavior and mood. Explained is that with depression, activities once pleasurable are often stopped which may heighten symptomatology. Likewise, previously pleasurable activities may no longer be pleasurable. Interventionists also discuss sharing depressive symptoms with physicians (if medication may be necessary or for future consideration) and strategies to use when physicians are of a different race. Interventionists introduce deep breathing as a basic stress reduction technique and help participants identify stressful points in the day for its use. In sessions 2-3, interventionists engage participants in simple problem solving to resolve identified care management needs (medical, housing repairs, relocation needs, social, benefits/entitlements), and works on a care management plan involving coordination, referral and linkages to other services if necessary. Other stress reduction techniques (counting, music) are introduced to provide additional tools. In sessions 4-5, interventionists continue addressing care management needs and introduce behavioral activation. Interventionists review daily routines and help participants select a goal and specific activity to add pleasure and personal satisfaction. Active problem solving and motivational interviewing techniques help participants achieve identified activity goals. Potential barriers in carrying out activities are identified and solutions derived which may also require care management (arranging transportation) to engage in desired activities. Sessions 6-8 involve reinforcement of activity engagement, identification of new activity goals and specific steps to achieve them. In sessions 9-10, interventionists review and reinforce all techniques, help participants identify and use behavioral activation strategies, and obtain closure. Approximately 6-7 sessions are devoted to behavioral activation.
Wait-list control group
The control group does not receive any study-related contact following the baseline interview. After completion of the 4-month reassessment, participants in this group receive the entire BTB intervention as delivered to the initial treatment group.
We based our sample size calculation on the following assumptions: a) one primary outcome (severity of depressive symptoms); b) treatment effect sizes reported in previously published intervention studies testing one or more components of the intervention we employ in BTB (effect sizes range from .70 at 6 months and .35 at 12 months in one trial16 to .40 at 3 and 6 months and .60 at 12 months for another trial14); c) ability to detect a medium effect size of Cohen's d = 0.40; and d) type I error rate of .05 for the single primary hypothesis. A larger effect size may not be plausible, whereas a smaller effect size brings us at or near levels where the study could have statistical but not clinical significance. In order for this brief, targeted intervention to be worth implementing in a service context, it must yield more than trivial effects. Since our planned analyses are primarily analysis of covariance, we calculated the power of a two-sample t-test. Because we will use analysis of covariance to adjust for baseline values, we expect to have adequate precision and power. To attain 90% power for a two-sided alternative hypothesis using a t-test to compare the two treatment groups with respect to 4-month values require 172 study participants for an effect size of .40. We planned to recruit an additional 20 participants for a total of 192 participants (96 per group). The additional study participants allowed for a conservative estimated attrition rate of 10%. To obtain the necessary sample size of 172, we needed to enroll 192 study participants. The statistical power for secondary analyses is the same as for the primary; 90% for a 0.40 effect size with 5% two-tailed tests. Although we did not conduct power calculations for exploratory aims, these analyses are important in order to refine research questions and hypotheses worthy of future consideration, refinement of the intervention and its dissemination.
The focus of the primary analysis is to examine the effect of the intervention on symptom severity based on the principle of "intention to treat." That is, all subjects assigned to the initial treatment group will be part of the primary analyses regardless of the actual number of intervention contacts received. This approach tests the effects of the intervention without accounting for the extent to which study participants actually receive the intervention. It effectively penalizes the intervention if subjects are not willing to receive it or if subjects receive different doses or exposure to treatment. We will calculate adjusted mean differences in treatment effects on the outcome measure at 4-months using analysis of covariance. We will include as covariates the baseline value of the outcome measure and the stratification variable (recruitment source). Other background characteristics may also be used as covariates if large or statistically significant differences between initial treatment and wait-list control group participants are found for those variables in analyses conducted prior to any efficacy analyses. Prior to conducting the ANCOVA, we will test the normality assumption for the dependent measure by examining the distribution of the residuals. If the residual distribution is skewed, we will use a transformation of the data.
For purposes of reporting results, we will also compute proportions improved. We will use two definitions of improvement: any reduction in PHQ-9; a reduction sufficient to move the patient at least one severity category lower. Note that we do not have to worry about the no symptom (0-4) at baseline group in this definition, as that group is not eligible. The comparison of treatment to wait-list control will be with odds ratio and 95% confidence intervals. We will also examine clinical significance by conducting subsidiary analyses that examine the proportion of individuals in treatment compared to controls who change diagnostic categories. Finally, in another set of analyses we will incorporate measures of the extent to which treatment has been received to evaluate relationship of dose/intensity to treatment outcome. This is for reporting purposes and we are not powering for these alternative endpoints.
As to our second aim, we seek to evaluate whether participants in the initial treatment group maintain reduced symptom presentation from 4 to 8 months. We view retention as a one-sided concept such that any beneficial effect of the intervention at four months is improved or at least maintained at 8 months. Lack of retention is then considered a non-trivial loss of benefit. In the language of clinical treatment trials, this is a "clinical equivalence" or "non-inferiority" hypothesis. Thus, our analyses for our second aim will focus on estimating the amount of 4-month benefit that is retained at 8 months for the treatment group only. We will use confidence intervals in reporting results in order to understand how much loss of benefit is consistent with the data we obtain. Given that the main outcome measure, PHQ-9, has cut off points related to DSM IV depression conditions (e.g., dysthymia, major depression), we will be able to evaluate loss of benefit and retention of effect within the treatment group based on clinically relevant features.
For exploratory aims, we seek to evaluate the mediational role of behavioral activation. We will use separate regressions in which the first regression will evaluate the relationship between 4 month values of the behavioral activation scale (the mediator) and symptom severity as measured by the PHQ-9 at 4 months. The second regression will evaluate the relationship between group assignment (the independent predictor variable) and symptom severity at 4-months (PHQ-9, the criterion). The third regression will evaluate the relationship of treatment assignment (predictor) to PHQ-9 severity scores (criterion) after entering behavioral activation values (mediator). To evaluate the strength of the mediation effects, we will compute the ratio of the unstandardized betas for treatment assignment (beta from regression III divided by beta from regression II) with values less than 1.00 indicating potential mediation. We will repeat these analyses using 8-month values of the PHQ-9. Another exploratory aim is to evaluate whether the intervention has a differential treatment effect based on a study participant's gender, age, or living arrangement (alone/with others), factors which have been shown to be differentially related to depression. To evaluate differential treatment effects at 4-months, we will use a similar analytic strategy as for the main effect model. ANCOVAs will be used in which the covariates will be the baseline value and specific participant characteristic (e.g., male/female; live with other or alone; age 60-64 versus 75+). We will then add an interaction term of group assignment by participant characteristic.
Economic evaluations of home support programs for depressive symptoms targeting older African Americans and which can be delivered by trained staff of a community-based agency are non-existent. Cost analyses are critical for evaluating the translational and implementation potential of such trials. We have calculated the cost of delivering BTB by senior center staff trained in the program (presented below); and we will calculate the cost effectiveness of BTB from a societal perspective that employs two relevant outcome measures: cost per quality adjusted life year (QALY; Euro-QOL 5D vs. health utility measures (HUI 2 and HUI3), and cost per reduction in depression symptoms (PHQ-9). Cost-effectiveness will be measured as the average incremental difference between cost of BTB and the control group divided by the difference in health utility (measured as QALYS) between treatment and control periods at 4 months, and at 8 months for the subsample of participants who receive BTB for the first 4 months and are observed for another 4 months. Univariate and probabilistic sensitivity analyses will be conducted to determine the incremental cost effectiveness of BTB at 8 months, and will aid in determining the robustness of the model when inpatient and outpatient medical and medication costs are varied. As funds for the cost analyses were obtained after the trial commenced, the cost analyses will involve a subset (n = 130/208) of participants for which we are able to collect QALY measures.
Recruitment and enrollment outcomes
Screening and enrollment results to date provide an indication of the feasibility of the BTB trial. A total of 17 senior center care managers were trained to use the PHQ-9 to screen for depression. Although screening and enrollment is closed for this trial, the screening protocol has become fully integrated into routine intake and reassessments across all senior center programs including IHSP.
A total of 703 older adults were initially screened over a 2 1/2 year recruitment period; 440 from IHSP; 263 from the community at-large (senior center members and elsewhere). Of 703, 390 (55%) scored positively (PHQ-9 ≥ 5) and were eligible for the second screen. Of 440 from IHSP, 137 (31%) were eligible for the second screen suggesting the importance of detection for this medically compromised population. Of 263 community at-large, 253 (96%) scored positively suggesting that media and outreach efforts successfully targeted individuals with depressive symptoms, individuals correctly self-identified as feeling depressed, and African Americans with depressive symptoms will volunteer for a non-pharmacological trial.
Of the initial 390 positive screens, 279 (72%) were successfully screened a second time of whom 192 (69%) were eligible and willing to participate in BTB. Another 16 individuals with initially negative PHQ-9 scores were screened 3 months later, became eligible, and enrolled in the trial for a total of 208 participants. As these participants contacted the study team prior to meeting the targeted sample size, they were included in the trial.
Participants are on average 69.5 years, mostly single, female, and with ≥ high school education. Over half (67.79%) report some to a lot of difficulty paying for basics, and have an average of 6.6 health conditions with more than a third having diabetes (42.8%), high blood pressure (78.8%), high cholesterol (58.2%), and arthritis (76.4%). Only 19.2% were on an anti-depressant medication and 16.8% on an anti-anxiety medication whereas 52.9% were on pain medications. The average second PHQ-9 score was 13 with 72% reporting moderate to severe symptoms. (Table )
Baseline Characteristics of BTB Sample (N = 208)
The total cost of BTB, reported in 2010 dollars, was obtained using a microcosting approach where total costs represent the sum of microcosts in 3 main categories: screening, intervention delivery, and program supervision. (Table ) To capture costs, senior center supervisors and interventionists kept detailed logs of time spent conducting BTB tasks. Though double screening over a two week period is required to confirm trial eligibility, costs for the first screening only are reported with the assumption that only one screening would be performed if BTB is translated for real world settings. Interventionists record real time in preparation, documentation and implementation for each session delivered in person or telephonically. Staff training time for interventionists is also captured and converted to costs based on application of hourly wage rates. Supervisors of interventionists at the senior center account for time spent with each interventionist in supervision. Also, cost of managing adverse events during the trial, such as suicidal ideations, is captured.
Costs for Delivery of Beat the Blues to Treatment Group to Date (n = 122)
Time costs are monetized by multiplying time required for each task for each participant, times the wage rate plus fringe benefit costs of employees performing the tasks. Wage rates and fringe benefit costs (healthcare, disability, life insurance) are obtained directly from employers (senior center and university). Non-time related costs (program materials and mileage) are also recorded. Material costs include study documentation forms and education print materials provided to BTB participants. Interventionist travel expenses to and from participants' homes are captured per visit, and monetized based on reimbursement at the current government rate ($0.55/mile).
The average screening cost is $2.63 per participant; total cost of BTB (screening plus home intervention) is $584.64 over 4 months, or $146.16 per participant, per month. Time required delivering BTB at home ($197.31) and labor and mileage costs associated with travel to homes ($211.03) were the major contributors to total model costs. Total cost of BTB compares favorably to brand name antidepressants such as Paxil (paroxetine), Zoloft (sertraline), or Cymbalta (duloxetine) which cost approximately $100-$300 for a one-month supply [30
]. Cost-effectiveness analyses of BTB from a societal perspective are currently underway.