Telephone-Delivered Cognitive Behavioral Therapy and Telephone-Delivered Nondirective Supportive Therapy for Rural Older Adults With Generalized Anxiety Disorder: A Randomized Clinical Trial

Importance  Generalized anxiety disorder (GAD) is common in older adults; however, access to treatment may be limited, particularly in rural areas.

Objective  To examine the effects of telephone-delivered cognitive behavioral therapy (CBT) compared with telephone-delivered nondirective supportive therapy (NST) in rural older adults with GAD.

Design, Setting, and Participants  Randomized clinical trial in the participants’ homes of 141 adults aged 60 years and older with a principal or coprincipal diagnosis of GAD who were recruited between January 27, 2011, and October 22, 2013.

Interventions  Telephone-delivered CBT consisted of as many as 11 sessions (9 were required) focused on recognition of anxiety symptoms, relaxation, cognitive restructuring, the use of coping statements, problem solving, worry control, behavioral activation, exposure therapy, and relapse prevention, with optional chapters on sleep and pain. Telephone-delivered NST consisted of 10 sessions focused on providing a supportive atmosphere in which participants could share and discuss their feelings and did not provide any direct suggestions for coping.

Main Outcomes and Measures  Primary outcomes included interviewer-rated anxiety severity (Hamilton Anxiety Rating Scale) and self-reported worry severity (Penn State Worry Questionnaire–Abbreviated) measured at baseline, 2 months’ follow-up, and 4 months’ follow-up. Mood-specific secondary outcomes included self-reported GAD symptoms (GAD Scale 7 Item) measured at baseline and 4 months’ follow-up and depressive symptoms (Beck Depression Inventory) measured at baseline, 2 months’ follow-up, and 4 months’ follow-up. Among the 141 participants, 70 were randomized to receive CBT and 71 to receive NST.

Results  At 4 months’ follow-up, there was a significantly greater decline in worry severity among participants in the telephone-delivered CBT group (difference in improvement, −4.07; 95% CI, −6.26 to −1.87; P = .004) but no significant differences in general anxiety symptoms (difference in improvement, −1.52; 95% CI, −4.07 to 1.03; P = .24). At 4 months’ follow-up, there was a significantly greater decline in GAD symptoms (difference in improvement, −2.36; 95% CI, −4.00 to −0.72; P = .005) and depressive symptoms (difference in improvement, −3.23; 95% CI, −5.97 to −0.50; P = .02) among participants in the telephone-delivered CBT group.

Conclusions and Relevance  In this trial, telephone-delivered CBT was superior to telephone-delivered NST in reducing worry, GAD symptoms, and depressive symptoms in older adults with GAD.

Trial Registration  clinicaltrials.gov Identifier: NCT01259596.

Generalized anxiety disorder (GAD) is characterized by excessive and uncontrollable worry accompanied by restlessness, fatigue, poor concentration, irritability, muscle tension, and/or sleep disturbance.¹ Generalized anxiety disorder is one of the most common anxiety disorders in older adults.² It is associated with poor quality of life,³ increased health care utilization,³ impaired memory,⁴ and potentially increased morbidity and mortality.^5-7

Medications and psychotherapy are the primary treatments for GAD. Cognitive behavioral therapy (CBT) has been found to be superior to minimal contact or usual care for late-life GAD.^8-11 Although pharmacological treatments have demonstrated some success in treating GAD,^12,13 they are associated with some potentially serious adverse effects for older adults^14,15 and alternatives to pharmacotherapy are needed. Further, many older adults prefer psychotherapy to pharmacotherapy for the treatment of anxiety.¹⁶

Quiz Ref IDThere are a number of barriers that older adults face, particularly those who live in rural areas. Mobility and transportation limitations can make travel to a professional’s office difficult.¹⁷ Older adults are also less likely to receive psychotherapy if they live in a county without a mental health professional,¹⁸ which is more of a problem for rural residents. Further, professionals specializing in late-life mental health disorders are even less common in rural areas. Thus, alternate methods of providing treatment may increase mental health care utilization by this underserved population.

In the current study, we adapted CBT for administration by telephone in an effort to overcome barriers to care while still providing an evidence-based treatment. We compared telephone-delivered CBT (CBT-T) with telephone-delivered nondirective supportive therapy (NST-T), a structurally equivalent comparison group¹⁹ with similar levels of outcome expectations and credibility.^20,21 Thus, this design allowed for a comparison between the gold standard for anxiety disorders and a commonly available type of psychotherapy in clinical practice. We compared the effects of CBT-T and NST-T on anxiety, worry, GAD symptoms, and depressive symptoms at 4 months’ postrandomization. We hypothesized that participants who received CBT-T would demonstrate greater declines in symptoms than participants who received NST-T.

Participants were adults aged 60 years and older with a principal/coprincipal diagnosis of GAD based on the Structured Clinical Interview for DSM-IV²² and lived in 1 of 41 rural counties in North Carolina with an urban population of less than 20 000 people. Exclusion criteria included current psychotherapy, active alcohol/substance abuse, dementia, or global cognitive impairment based on the Telephone Interview for Cognitive Status–Modified²³; psychotic symptoms, active suicidal ideation with plan and intent, change in psychotropic medications in the last month, bipolar disorder, and hearing loss that would prevent a person from participating in telephone sessions. This study was approved by the Wake Forest School of Medicine Institutional Review Board. Participants provided verbal informed consent to be screened and written informed consent to undergo the diagnostic interview. The full study protocol can be found in the trial protocol in Supplement 1.

The procedures of this study have already been described in detail.²⁴ A commercial mailing company was used to mail study flyers to older adults living in rural counties. Interested participants called and completed a brief screener. Then, they completed a diagnostic interview by telephone and were mailed self-reporting measures. Eligible participants were randomized to 1 of 2 treatments and 1 of 3 therapists. Randomization was stratified on baseline depression diagnosis (lifetime major depression/dysthymia vs no depression) and psychotropic medication use (yes/no) and was administered via a secure web-based data management system using a permuted block algorithm and random block lengths. Randomization procedures were performed by staff who were not involved in the assessments.

All study assessments were completed either by telephone or mail. The baseline assessment consisted of self-reported demographic information (age, sex, race/ethnicity, marital status, education, income, employment status, and living status), health-related information (comorbidities, medications, and smoking status), and primary and secondary outcomes. Race/ethnicity data were collected for future moderator analyses. With the exception of the health-related information provided during the diagnostic interview, this information was completed via self-reported measures. A brief follow-up assessment was conducted 2 months after randomization. At 4 months after randomization, all measures except demographic characteristics were reassessed. The primary hypotheses of this study were based on 4 months’ postrandomization data. Type I error was allocated at 0.025 for each outcome so that if either end point was significant, the trial would be considered to have positive results. Only anxiety and mood-related outcomes are presented. Interviewers were masked to participants’ treatment groups.

Anxiety symptoms were assessed with the Structured Interview Guide for the Hamilton Anxiety Rating Scale (HAMA),^25,26 a 14-item interviewer-rated measure of general anxiety. Ratings were made on a 5-point scale, ranging from 0 (none) to 4 (very severe). The HAMA has been validated in samples of older adults with GAD and demonstrates good interrater reliability (r = 0.81-0.85).^11,27,28 A total of 10% of HAMA interviews were randomly selected and rated by an assessor masked to condition and otherwise not involved with the study (intraclass correlation coefficient = 0.95).

Worry was assessed with the Penn State Worry Questionnaire–Abbreviated (PSWQ-A),^29,30 an 8-item self-reported measure of the frequency and intensity of worry. Participants rated each item on a 5-point scale and responses were summed, with higher scores indicating greater worry. The PSWQ-A has better test-retest reliability and comparable validity as the full-length version.^29,31 The internal consistency of the PSWQ-A in our previous study was 0.86.³²

The GAD Scale 7 Item³³ is a self-reported measure of DSM-IV symptoms of GAD. Participants rated 7 questions on a 4-point Likert scale and responses were summed to create a total score. It has good internal consistency (α = 0.89-0.92) and test-retest reliability (intraclass correlation coefficient = 0.83).^33,34

The Beck Depression Inventory (BDI)³⁵ is a 21-item measure of depressive symptoms. Responses were summed, with higher scores indicating greater depressive symptoms. The BDI has good psychometric properties in samples of older adults with GAD.³⁶

The Expectancy Rating Scale³⁷ is a 4-item questionnaire that assesses beliefs in how logical the treatment seems, confidence in undergoing treatment and recommending it to others, and expectations for success. It was administered after the first session.

The Client Satisfaction Questionnaire³⁸ is an 8-item questionnaire that assesses patient satisfaction with treatment. It was administered at the 4-month assessment.

Treatment was delivered by 2 graduate-level social workers and 1 postdoctoral clinical psychologist. Therapists delivered both treatments and were masked to assessment data. They received weekly supervision for CBT-T (G.A.B.) and NST-T (S.C.D.). Participants randomized to CBT-T received 9 to 11 weekly 50-minute telephone therapy sessions that corresponded with the CBT workbook. Each workbook chapter³⁹ included CBT techniques that have demonstrated efficacy in treating adults^40,41 and older adults with GAD.^8,11,42 Chapters addressed recognition of anxiety symptoms, relaxation, cognitive restructuring, the use of coping statements, problem solving, worry control, behavioral activation, exposure therapy, and relapse prevention (with optional chapters on sleep and pain). Participants randomized to NST-T received 10 weekly 50-minute therapy sessions. Telephone-directed nondirective supportive therapy was based on the Borkovec et al protocol^20,43 and provided a high-quality therapeutic relationship with a warm, genuine, and accepting atmosphere through the use of supportive and reflective communications; it did not provide advice, suggestions, or coping methods. Participants in both groups received 4 booster sessions at 2, 4, 8, and 12 weeks following completion of the weekly sessions.

All therapy sessions were audiotaped and 10% (1 session from each participant) were randomly selected to be rated by 2 postdoctoral-level clinicians not affiliated with the study. The sessions were evaluated for therapist adherence to the protocol and competence in delivering the interventions on a 9-point scale used in other late-life GAD psychotherapy studies.^8,10 Mean therapist ratings for adherence and competence were higher than 6.0 (good) for both treatments for all therapists, which was the a priori requisite.

Assuming that each primary outcome would be tested at the 4-month assessment using a .03 2-sided significance level (ie, a Bonferroni correction), power calculations for analysis of covariance on the PSWQ-A outcome indicated the need for a total of 80 participants per treatment to attain 90% power, 70 participants per treatment to provide 85% power, and 60 participants per treatment to attain 78% power.²⁴ We planned to randomize 88 participants per treatment to account for an expected dropout rate of approximately 10%. In October 2013, conditional power calculations based on having 80 evaluable participants per treatment were presented to the data safety monitoring board to illustrate the effect on power of stopping recruitment early. At that time, the total number of participants with 4-month assessment measurements was projected to be 120 and it was determined that further randomization, even to a total of 160 evaluable participants per treatment, would most likely not have an effect on conclusions for the 2 primary outcomes. Thus, randomization was stopped and follow-up continued (see eAppendix 1 in Supplement 2 for details).

All analyses were performed using SAS, version 9.4 (SAS Institute Inc). Baseline characteristics were summarized overall and by treatment using means and percentages. The percentage of therapy sessions attended was calculated for each person, assuming a maximum of 10 sessions per person (ie, some CBT-T participants had an optional 11th session but were only credited with 10 sessions for the computation). The mean percentage of sessions attended was then compared between treatments.

The prespecified analysis plan included all randomized participants as allocated to their original treatment group and used maximum likelihood estimation within a population-averaged repeated-measures analysis of covariance, with an unstructured covariance matrix to account for the fact that the multiple follow-up measurements from participants were not independent (PROC MIXED of SAS was used). This approach is consistent with the principle of intent to treat when at least 1 follow-up measure is obtained prior to dropout.⁴⁴ The prespecified time for comparison of the primary outcomes between treatments was the 4-month visit and the hypothesis test and was performed as a linear contrast within the analysis of covariance model using a 2-sided .02 significance level (see eAppendix 1 in Supplement 2 for explanation of significance level based on interim analysis). The analysis of covariance model for the primary outcomes contained terms for the baseline value of the outcome, therapist (a factor to which participants were randomized), lifetime depressive disorder (used to stratify randomization), use of psychotropic medications at baseline (used to stratify randomization), treatment effect, time effect, and time-by-treatment interaction term. Prespecified secondary outcomes included the BDI and GAD-7 and were analyzed using the same approach as specified for the primary outcomes. Nominal significance levels were reported for these 2 secondary outcomes and they were tested at the .05 level. Under independence between these outcomes, the probability of type I error was 0.0975 but less than 0.075 if the outcomes were positively correlated at 0.4 or larger.⁴⁵ There were no prespecified subgroup analyses. Because some participants did not have any follow-up measurements, a full intent-to-treat sensitivity analysis used multiple imputation to explore the effect of missing outcomes on the overall conclusions for each outcome (HAMA, PSWQ-A, GAD-7, and BDI). A description of this analysis and the findings are included in eAppendix 2 in Supplement 2. Lastly, treatment response rates and the proportions without 4-month measurements were compared using the Fisher exact or χ² tests and a Wilcoxon test was used to compare levels of treatment expectancies and satisfaction between treatments.

Quiz Ref IDRecruitment occurred between January 27, 2011, and October 22, 2013. The initial recruitment goal was 176 participants; however, recruitment was slower than anticipated. Recruitment was stopped when 141 participants were randomized (CBT-T, n = 70; NST-T, n = 71). Figure 1 displays participant flow from randomization through study follow-up and provides details on dropout rates. Almost 75% of participants in CBT-T (52/70) completed the required 9 sessions and more than 81% (58/71) of participants in NST-T completed the required 10 sessions. Two participants died and 1 participant became seriously ill prior to the first session. Data completion rates for the outcome measures ranged from 85.8% to 89.9% at the 2-month assessment and 79.7% to 90.0% at the 4-month assessment (percentages based on the total number randomized). Fifteen of 70 participants in CBT-T (21.4%) and 8 of 71 participants in NST-T (11.3%) were missing at least 1 of the primary outcomes at the 4-month assessment (P = .12 by Fisher exact test). Ten of 70 participants in CBT-T (14.2%) and 3 of 71 participants in NST-T (4.2%) withdrew consent at some point prior to the 4-month visit (P = .05 by Fisher exact test).

The Table presents the demographic and health-related characteristics and baseline outcome values.

Quiz Ref IDFigure 2 displays the results for the primary outcomes. There was a significant decline in anxiety symptoms (HAMA) among participants in both treatments (CBT-T: −7.04; 95% CI, −8.88 to −5.21; NST-T: −5.53; 95% CI, −7.28 to −3.77) but there was no significant effect of treatment on general anxiety symptoms (difference in improvement, −1.52; 95% CI, −4.07 to 1.03; P = .24). Similarly, there was a significant decline in worry (PSWQ-A) among participants in both treatments (CBT-T: −9.44; 95% CI, −11.02 to −7.86; NST-T: −5.37; 95% CI, −6.89 to −3.85). However, there was a significantly greater decline in worry among participants in CBT-T (difference in improvement, −4.07; 95% CI, −6.26 to −1.87; P = .004).

Figure 3 displays the results for the secondary outcomes. There was a significant decline in self-reported GAD symptoms (GAD-7) among participants in both treatments (CBT-T: −5.63; 95% CI, −6.82 to −4.44; NST-T: −3.27; 95% CI, −4.40 to −2.14), with a significantly greater decline among participants in CBT-T (difference in improvement, −2.36; 95% CI, −4.00 to −0.72; P = .005). Similarly, depressive symptoms (BDI) declined among all participants (CBT-T: −10.77; 95% CI, −12.73 to −8.81; NST-T: −7.54; 95% CI, −9.44 to −5.64). Similar to GAD symptoms, there was a significantly greater decline in depressive symptoms among participants in CBT-T (difference in improvement, −3.23; 95% CI, −5.97 to −0.50; P = .02).

Quiz Ref IDA meaningful response to treatment has been defined as a 5.5-point decrease in PSWQ-A scores from baseline to posttreatment.^10,46,47 A significantly greater proportion of participants in CBT-T experienced a reduction in PSWQ-A scores by at least 5.5 points (72.4%) compared with participants in NST-T (42.9%; P = .001).

Quiz Ref IDThere were no significant differences between the means (SD) of treatments with respect to their expectations of the treatment (CBT-T, 29.6 [7.51]; NST-T, 29.4 [7.45]; P = .09). However, mean (SD) satisfaction was significantly higher among participants in CBT-T (27.9 [4.48]) compared with participants in NST-T (25.2 [5.86]; P = .01).

A total of 8 participants (4 in each condition) experienced a greater than 1 SD increase in PSWQ-A scores (3 at the 2-month assessment and 1 at the 4-month assessment) or in BDI scores (2 at the 2-month assessment, 1 at the 4-month assessment, and 1 at both the 2-month and 4-month assessments). These participants were assessed for immediate need for psychiatric treatment. Referrals for additional therapy or medication management were provided if needed.

We found that both treatments reduced symptoms of worry, depression, and GAD. However, CBT-T was superior to NST-T, resulting in a significantly greater reduction in these symptoms. To our knowledge, this is the first study of a telephone-delivered intervention for the treatment of late-life GAD as well as the largest head-to-head comparison of individual CBT with an active psychotherapy comparator.

Telephone-delivered psychotherapy is 1 way to overcome some barriers to mental health treatment that rural older adults face. Although we did not compare telephone-delivered psychotherapy with face-to-face treatment, research suggests that telephone-delivered treatment may be equally efficacious and yield lower attrition rates.⁴⁸ The effect of telephone-delivered CBT on worry in the current study (prepost change/baseline SD, 1.37) is comparable with those found in other studies of late-life GAD in which CBT was delivered in a face-to-face format.^{8,11,42,49,50} Further, our findings suggest that telephone-delivered psychotherapy is well liked. First, scores on the satisfaction measure were high. Second, adherence was high in this study; more than 75% of the sample completed at least 9 sessions. Third, attrition was lower in the current study (11.3%) than comparable studies of psychotherapy for late-life GAD (attrition rates for similar studies range from 13.5% to 33%).^10,42

Few studies have focused specifically on psychotherapy for treating late-life GAD and only 1 study⁴² has compared CBT with another type of active psychotherapy. This randomized design allowed for a rigorous test of CBT, controlling for common therapeutic elements across psychotherapies. Stanley and colleagues⁴² compared CBT with supportive therapy delivered in a group format (N = 48) and found no significant differences between the 2 treatments. Potential explanations for the difference include sample size (only 31 participants completed their treatment) and the CBT intervention (theirs consisted of the following 3 components: relaxation, cognitive therapy, and exposure while the current treatment included 8 to 10 different topics). Although we found that both treatments improved symptoms, CBT-T was significantly better than NST-T at reducing worry, GAD, and depressive symptoms.

There were a number of strengths of this study, including a rigorous test of CBT by comparing it with an active treatment in a randomized clinical trial, adaptation of treatment to meet the needs of a rural population, high treatment adherence and integrity, and low attrition rates. To address any potential effect of depression on treatment, randomization was stratified by the presence of major depression or dysthymia. Notably, CBT-T demonstrated superiority in treating comorbid depression symptoms. Similarly, the use of psychotropic medications, another possible confounding variable, was also used as a stratification variable. Nonetheless, there were some limitations. Although scores on the HAMA declined, there were no significant differences between the 2 conditions. One possibility may be owing to poor reliability of this interviewer-rated measure. However, 10% of HAMA interviews were randomly selected and rated by a masked reviewer. The intraclass correlation coefficient was 0.95, suggesting that interviewer variance was not responsible for the lack of difference. An alternative explanation is that while it is superior at reducing the worry symptoms specific to GAD, CBT-T is less effective for somatic symptoms of anxiety. Further, biased responding by participants on self-reported measures cannot be ruled out; however, there was no reason to believe it would be differential among randomized groups. Another limitation was the largely white well-educated sample that potentially limited generalizability of the findings, particularly to those with lower educational or literacy levels. Finally, a decrease in symptoms from baseline to follow-up may have reflected regression to the mean.

Among rural older adults diagnosed as having GAD, both CBT-T and NST-T are associated with reductions in anxiety, worry, GAD, and depressive symptoms. However, CBT-T was superior to NST-T in reducing worry, GAD, and depressive symptoms. There was no differential effect of treatment on anxiety symptoms.

Corresponding Author: Gretchen A. Brenes, PhD, Department of Psychiatry and Behavioral Medicine, Wake Forest School of Medicine, Medical Center Boulevard, Winston-Salem, NC 27157 (gbrenes@wakehealth.edu).

Submitted for Publication: March 18, 2015; final revision received May 27, 2015; accepted May 28, 2015.

Published Online: August 5, 2015. doi:10.1001/jamapsychiatry.2015.1154.

Author Contributions: Dr Brenes had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Brenes, Miller.

Acquisition, analysis, or interpretation of data: All authors.

Drafting of the manuscript: Brenes, Danhauer, Lyles, Miller.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Brenes, Hogan, Miller.

Obtained funding: Brenes, Miller.

Administrative, technical, or material support: Brenes, Danhauer, Lyles, Miller.

Study supervision: Brenes, Danhauer, Miller.

Conflict of Interest Disclosures: None reported.

Funding/Support: This work was funded by grant R01 MH083664 from the National Institute of Mental Health (Dr Brenes).

Role of the Funder/Sponsor: The funder had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Disclaimer: The contents of this publication are solely the responsibility of the authors and do not necessarily represent the official views of the National Institute of Mental Health.

Additional Contributions: We thank Elizabeth Arnold, PhD, Wake Forest School of Medicine, and Melinda A. Stanley, PhD, Baylor College of Medicine, for providing treatment adherence and integrity ratings for the therapy sessions. We also thank those who served on the data safety monitoring board, including Patricia Arean, PhD, University of California at San Francisco School of Medicine, John Preisser, PhD, University of North Carolina Gillings School of Global Public Health, and Julie Wetherell, PhD, University of California at San Diego School of Medicine, chair. All contributors received compensation for their work.