|Home | About | Journals | Submit | Contact Us | Français|
Despite a marked increase in persons seeking help for anxiety disorders, the care provided may not be evidence-based, especially when delivered by non-specialists. Since anxiety disorders are most often treated in primary care, quality improvement interventions are needed there.
A randomized controlled trial of a collaborative care effectiveness intervention for anxiety disorders.
Approximately 1040 adult primary care patients with one of four anxiety disorders (generalized anxiety disorder, panic disorder, posttraumatic stress disorder, or social anxiety disorder), recruited from four national sites.
Anxiety clinical specialists deliver education and behavioral activation to intervention patients and monitor their symptoms. Intervention patients choose cognitive behavioral therapy, anti-anxiety medications, or both, in a “stepped care” treatment that varies according to clinical need. Control patients receive usual care from their primary care clinician. CALM's innovations include the flexibility to treat any one of four anxiety disorders, co-occurring depression, and/or alcohol abuse; its use of on-site clinicians to conduct initial assessments, and its computer-assisted psychotherapy delivery.
Anxiety symptoms, functioning, satisfaction with care, and health care utilization are assessed at 6-month intervals.
CALM was designed for clinical effectiveness and easy dissemination in a variety of primary care settings.
About 11% of the US population will suffer from an anxiety disorder each year and almost 29% will experience an anxiety disorder at some point in their lives . Despite a marked increase in the proportion of individuals seeking help for anxiety disorders in the last ten years , their care may not be evidence-based, especially when provided by non-specialists. Since persons with anxiety disorders are most often treated in primary care settings, quality improvement interventions within those settings are needed.
One approach to improve mental health care in primary care settings is the “collaborative care” model [4-7], closely patterned on the “chronic disease model” . Various forms of the collaborative care model have been tested but some features are common to all.[9-12] Patients typically remain under the care of their primary care provider who is assisted by a care manager, usually a masters-level clinician (e.g., nurse, social worker), working in consultation with a psychiatrist. In addition to expert consultation and redesign of clinical practice with care managers, collaborative care interventions contain other components useful in chronic disease management, including techniques to help patients manage their condition (patient education, psychotherapy, motivational enhancement, approaches to identify and reduce treatment barriers), and ongoing clinical monitoring of outcomes. Collaborative care interventions are “bundles” of practices that contribute to overall care delivery. These interventions have recently become more flexible by allowing patients a choice of treatments.
Most studies of collaborative care have been designed to assist primary care physicians treat depression.[11-14] Our previous work, Collaborative Care for Anxiety and Panic (CCAP), is one example of a collaborative care intervention for panic disorder. Similar work has tested collaborative care models for treatment of panic disorder  and for treatment of both panic and generalized anxiety disorders in primary care . Collaborative care models are beginning to be used to improve the treatment of substance use disorders in primary care  and for management of depression within specialty medical clinics, such as HIV clinics (Pyne, JS, personal communication).
Many collaborative care models have been shown to be both clinically and cost effective, yet these models are not yet widely-used and are rarely sustained beyond the period of external research funding,[4-7,9,19,20] though this now may be changing.  Because of the many barriers to funding for and implementation of collaborative care, making these interventions as “user-friendly” as possible is key to their successful dissemination.
This paper describes the design and rationale of the Coordinated Anxiety Learning and Management (CALM) study, the largest randomized trial of collaborative care for anxiety disorders to date. CALM contains a number of innovations, including (1) treating any one of four anxiety disorders—panic (PD), generalized anxiety (GAD), post-traumatic stress disorder (PTSD), social anxiety (SAD)—in a single intervention; (2) accommodating multiple comorbidities, including depression, alcohol abuse, and chronic medical conditions; (3) emphasizing cognitive behavioral therapy (CBT), and providing innovative CBT training and delivery; and (4) using clinicians rather than research personnel to perform assessments. In addition, CALM incorporates newer, more flexible features initially used in the IMPACT study , including patient choice, “stepped care” (rather than a rigid treatment protocol), and real-time clinical monitoring through a structured web-site accessible to all key study personnel.
Based in part on feedback from primary care providers who wanted an optimally “user friendly” collaborative care model the CALM intervention was developed to maximize ease of dissemination in a variety of primary care settings lacking existing mental health expertise.
Patients are recruited and treated at four sites: Seattle, Los Angeles, San Diego, and Little Rock. Centralized data collection (telephone baseline and follow-up assessments) by the RAND Survey Research Group offers both efficiency and uniformity as data is collected across all sites by the same set of interviewers who are blind to the patients' status. Each of the four treatment sites will recruit 260 subjects from a variety of primary care clinics. Institutional Review Boards at all five institutions approved the protocol.
CALM is a prospective, longitudinal, randomized controlled trial of a collaborative care intervention that can provide treatment for any of four anxiety disorders: PD, GAD, PTSD, and SAD whose treatments are similar. Each can be treated with the same first-line medications, selective serotonin reuptake inhibitors (SSRI's), typically at similar dosages. Likewise CBT techniques such as psychoeducation, breathing retraining, cognitive restructuring, and exposure are similar in structure and format across these disorders, with the precise treatment tailored to the specifics of each anxiety disorder. The application of these cognitive and behavioral approaches differs across these four disorders primarily due to the variation in the cues that provoke anxiety.[24-26]  Obsessive compulsive disorder was not included because its recommended treatment regimen, especially pharmacologic, is considerably more complex.
The comparison condition to CALM is usual care (UC) treatment conducted under naturalistic conditions. UC is whatever treatment the provider would use in everyday practice. This typically entails primary care physician pharmacotherapy for the majority of patients and referral to outpatient mental health specialists for a smaller proportion. No restrictions were placed on providers' usual care practices. They use whatever treatment modalities they prefer or refer patients out as deemed appropriate. Providers are informed of the research diagnosis regardless of patient assignment to the CALM intervention or UC. Provision of diagnostic information has not improved outcome appreciably in the UC arm of previous collaborative care studies.[4-6,15]
Each of the four sites selected clinics in their geographic area to participate. Candidate clinics were evaluated and purposively selected based on a number of considerations, including provider interest, space availability, size and diversity of the patient population, and insurance mix. We attempted to maximize including patients from a variety of racial/ethnic groups and income levels, and with a variety of insurance sources (public and private) and to include a variety of types of primary care clinics. For example, one clinic (in San Diego) was chosen because it serves primarily Spanish-speaking patients. Varied provider models across clinics meant that CALM negotiated varying agreements with each site to reimburse them for office space and in some cases for additional administrative overhead costs. Together, these clinics serve over 350,000 individuals each year with more than 780,000 visits (see Table 1). Only 7 of the 12 clinics routinely have some form of in-house mental health services on site, primarily consultative. At each site a clinician “champion” serves as liaison to the CALM study.
ACSs conduct initial assessments of referred patients and manage and deliver treatment (with the support of the primary care provider) to those patients randomized to the CALM intervention. We sought individuals with patient-care experience, not necessarily in mental health, and some exposure to the primary care setting. To more closely approximate conditions faced by most primary care clinics, we did not seek individuals with expertise in anxiety management or CBT. Table 2 shows the ACS characteristics. Ranging in age from 25 to 59, most study ACSs are women with master's degrees in social work or nursing. 8 of 14 had any experience with mental health; half had previous experience with medication management, and 4 of 14 had some exposure to CBT. Most ACSs were selected by study personnel and are employed by the study site academic institution but some were selected and employed by the health care provider organization operating the clinic.
ACSs conduct initial assessments, a departure from other collaborative care studies that used research personnel to gather this information. This approach was chosen to facilitate potential dissemination outside a research setting. The ACSs also provide education about anxiety, prepare patients to make treatment choices, address barriers to treatment, assist the prescribing physician with medication management, and deliver computer-assisted CBT. Once trained, ACSs are supervised by study clinicians (psychiatrists and PhD psychologists) in weekly sessions within site and in monthly cross-site supervision conference calls. In addition, local psychiatrists and psychologists participate in a monthly conference call (without the ACSs) to address medication recommendations and supervision of CBT across sites.
Our “facilitated referral” approach uses multiple strategies to recruit subjects. Primary care providers and clinic nursing staff directly refer potential subjects, and we actively publicize the study within each clinic. We sponsor provider lunches where we describe the study and emphasize the benefits for patients. A simple five-question screener with study contact information was provided to clinics. These screeners could be posted in waiting rooms or examination rooms, handed to waiting patients, or administered directly by the physician. We continue to remind busy clinic personnel and providers about the study by distributing promotional materials e.g., bookmarks, t-shirts, and mugs.
An eligible subject must be a patient at one of the participating clinics, be at least 18 years old, meet DSM-IV criteria for GAD, PD, SAD, or PTSD, be willing to participate in CALM, and be able to provide written, informed consent. Exclusion criteria were minimal and were intended to exclude persons who would not likely benefit from the intervention or for whom the intervention could be risky. They included serious alcohol or drug use (specifically, alcohol or marijuana dependence or any other drug abuse or dependence, including methadone), unstable medical conditions, marked cognitive impairment, active suicidal intent or plan, psychosis, or bipolar I disorder. Subjects already receiving ongoing medication management or CBT were excluded. Finally, persons without routine access to a telephone, or who could not speak English or Spanish were excluded.
ACSs ascertained eligibility in face-to-face interviews using the Mini International Neuropsychiatric Interview (MINI) (Version 5.0)  sections on PD, SAD, PTSD, GAD, and Major Depression and other modules essential for determining ineligibility (Suicidality, Manic/Hypomanic Episode, Alcohol Abuse/Dependence, Drug Abuse/Dependence, and Psychotic Disorders). We selected the MINI as the simplest, most clinically relevant and easiest to learn interview, anticipating that, if this intervention were disseminated widely, eventual “real world” patient evaluations would likely be conducted by a therapist/case manager who might lack extensive education in mental health and need some form of structured interview. The MINI is the shortest instrument currently available that can diagnose four anxiety disorders along with relevant comorbidities such as major depression and substance abuse, and important study exclusions such as psychosis and bipolar disorder. The MINI takes about 45 minutes to administer. In addition to meeting criteria for disorder, patients must score at least 8 (moderate but clinically significant anxiety symptoms on a scale ranging from 0 to 20) on the Overall Anxiety Severity And Impairment Scale (OASIS). The Alcohol Use Disorders Identification Test (AUDIT) is used to screen for alcohol dependence and simple queries about drug use are employed to screen for drug use.
We considered randomizing clinics rather than randomizing patients. However, this approach would significantly reduce statistical powers, and we , and others [4,5,9], have successfully utilized randomization of patients without incurring any significant “spillover” effects where primary care physicians treat patients in both intervention and UC arms. RAND performed randomization procedures for eligible patients at all sites using a computerized random number generator. In this way both the ACS who enrolled the patient and the interviewer who conducted the baseline interview are blind to assignment. We use a stratified permuted block randomization. Strata are defined by the clinics crossed with depression status for a total of 32 strata. Within every stratum we randomize patients using a permuted block design with a fixed block size. The block size is masked to all the study members with the exception of the statistician and programmer who generate the random allocation. Keeping the block size hidden reduces the likelihood of protocol subversions or indirect selection bias. The choice of the stratified permuted block design was dictated by several factors. Primarily, we wanted to ensure balance of conditions (intervention and UC) within every stratum to give us the flexibility to compare at the clinic or stratum level if needed.
CALM was modeled after Unutzer's Improving Mood: Promoting Access to Collaborative Treatment for Late-life Depression (IMPACT) intervention which utilized a “stepped care” approach . Initial courses of treatment, either medication or CBT, require approximately 10-12 weeks. Patients meeting criteria for multiple anxiety disorders choose the most disabling or distressing disorder for treatment. (There is substantial evidence from both the CBT and psychopharmacology literature that successful treatment of one anxiety disorder generalizes to other anxiety disorders as well as mood disorders). Treatment continues until the patient shows sufficient improvement, either by “stepping up” (increasing) the current treatment or “stepping over” to or adding a different treatment modality. In keeping with recent standards in medication trials and some psychotherapy studies, the aim of treatment is clinical remission, defined as an OASIS score of 5 or lower indicating mild or no symptoms. While depression is not the focus of treatment, subjects are also rated on a shortened version of the PHQ-9, the PHQ-3  (J. Unutzer, personal communication) to monitor progress with frequently co-morbid depressive symptoms. Patients are monitored for treatment response and additional steps of care, pharmacotherapy or CBT, or a combination of the two, are provided as needed. In addition to brief assessments at each patient contact, the ACS assesses patient response formally at 3-month intervals. For patients not showing evidence of significant improvement the psychiatrist arranges an inperson consultation with the patient and reviews the case with the team. Psychologist consultation is available when CBT needs to be altered or added. Once patients respond, the ACS follows them with monthly telephone calls. If symptoms re-emerge within the first nine months of the study, the ACS can refer the patient back to the next step of care. In some cases, psychiatric assessment could reveal that a patient with symptomatic improvement, who had not yet reached remission but was close, would be unlikely to improve further with the treatments available (e.g., chronic ongoing stress limiting complete recovery; severe childhood abuse requiring a more prolonged or complex treatment course; poor engagement with care). These subjects enter the monthly follow-up phase and can re-enter more intensive treatment if their circumstances change. To clarify more complicated or puzzling diagnostic presentations (e.g., information given by the patient varies from his or her response to MINI questions), study clinicians (psychiatrists or PhD psychologists) discuss these cases by phone or email. All new patients are discussed in weekly supervision. The ACS interacts regularly with the primary care physician both face-to-face and via written communication. The primary care physician remains in charge of the patient's treatment, and, as the prescriber of all psychotropic medication, is the principal study psychopharmacologist.
Cognitive behavioral therapy, a short term, directive therapy derived from learning theory and the cognitive science of emotion, has been shown effective for 60 to 100% of persons who receive a course of treatment and attrition rates are generally low -- between 10 and 25%.[24-26]. General strategies for CBT treatment (i.e., psychoeducation, self-monitoring, cognitive restructuring, somatic exercises, and exposure to fear related stimuli) do not vary across anxiety disorders but are tailored to the specific triggers and manifestations or symptoms of the given disorder being treated. For example, exposure treatment entails exposure to feared situations in panic disorder and to social situations for social anxiety, reminders of trauma in PTSD, and behavior modification of overly cautious behaviors in GAD.
Although CBT is a known evidence-based treatment it has not been disseminated widely, partly because training for clinicians is not available in many areas. One way to enhance dissemination is through computer-assisted programs delivered with a therapist present. Computerizing CBT also facilitates iterative learning of CBT strategies by novice therapists, who do not have to remember all the specific details and sequence of steps.
Since no computer-assisted CBT approach existed for all CALM disorders, we created this de novo. CALM's computer assisted program, used by the ACS and patient together on a stand-alone computer at the primary care clinic, includes psychoeducational materials and instructions for skill practice and exposure. It demonstrates techniques with patient-actors and is interactive so data may be entered by the ACS, and a personalized workbook created that includes homework assignments. The ACS directs the patient to the appropriate sections of the computerized treatment, assesses patient understanding of the material, and assists in applying CBT principles and exercises to the patient's idiosyncratic thoughts and behaviors. The ACS is present at all times and integrally involved in helping the patient learn the CALM program. The computerized program includes five basic modules (Getting Started, Calm Education, Calm Recording, Calm Breathing, and Keep Going) and three other modules (Calm Thinking, Calm Living, and Calm Feeling), tailored to each anxiety disorder through branching mechanisms. With some variability, patients require about eight 60-minute sessions to complete the program. Some data gathered during the course of treatment sessions (e.g., homework adherence), is stored in the personal computer and periodically downloaded for analysis at the conclusion of the CALM study.
Based on the limited evidence-base, an SSRI or SNRI is usually selected as initial pharmacotherapy. The choice is individualized, based on the patient's history of use and non-response, availability of a generic version, side-effect profile, and potential drug interactions. Initial dosing is low with gradual upward titration to maximally tolerated doses over 4-6 weeks, with the aim of a treatment “trial” being a minimum of 10-12 weeks. Only if these agents had been tried and found ineffective in the past, would a GABAergic strategy (benzodiazepine or gabapentin) be considered early on. Keeping in mind there is no evidence base for further adaptive medication strategies, at 12 weeks, would be switched from the initial SSRI to another SSRI or an SNRI, or a GABAergic agent. Partial responders might receive a GABAergic agent added to their antidepressant. Additional options for poor responders would include addition of a second antidepressant such as mirtazapine, or a tricyclic antidepressant, with an even later option being addition of an atypical antipsychotic. In all cases, addition of CBT could substitute for further pharmacotherapy adjustments.
The ACSs read essential information (DSM-IV anxiety disorders and CBT principles) and attended six didactic half-day to full-day workshops addressing CBT in general, CBT in CALM, and the tailoring of CBT to each of the four anxiety disorders. Each workshop ended with a knowledge-base quiz. Workshops were interspersed with telephone-administered role plays in which the ACSs were given a set of 6-8 scenarios for each anxiety disorder that tapped the various aspects of CBT and were asked to role play delivery of each aspect. We evaluated the effectiveness of ACS training based on their performance with two to six training patients. Sessions were audio taped, reviewed by CBT supervisors and rated for proficiency (adherence and competence). ACSs had to demonstrate moderate or greater proficiency on at least 50% of their training cases.
ACSs also received a two-hour didactic on anti-anxiety medications and the medication algorithm for CALM. Strategies for explaining the rationale and effects of medications were reviewed, and protocols for monitoring medications (beneficial effects and adverse effects) were presented. The CALM manual also provides tables of medications and side effects, specific approaches to managing side-effects, and specific indications for choosing one type of medication vs. another. The didactic presentation was followed by two proficiency quizzes testing general approaches to managing common medication concerns (non-response, side effects) and knowledge of medication effects and side effects. The most important aspect of training, however, is weekly supervision of medication management, during which cases are reviewed and ACSs receive further training in selecting, monitoring, and evaluating the suitability for continuation or discontinuation, of various medications.
CALM's web-based tracking system was developed by Youlim Choi, Senior Web Programmer at the University of Washington, and was a modification of the system created for Unutzer's IMPACT study.[9,38] The web-based system offers several advantages over traditional paper-based records systems. It allows for real-time monitoring of recruitment, enrollment, diagnoses, ineligibilities, patient contact information, and continuous symptom assessments and can be accessed by all key study personnel to monitor the progress of the CALM study. RAND interviewers use the system for contact information, enrollment status, diagnosis, and status of assessments.
The web-based system supports ACSs by providing clinical reminders that facilitate treatment according to the intervention protocol. (It should not be confused with the computerized CBT system ACSs use to deliver the intervention to patients, which is not web-based). Screens are highly formatted to ensure recording of all necessary information. The note for the initial visit includes information on social history, medical history, family mental health history, current and past use of prescription psychiatric medications (dose, duration, side effects, efficacy) and current medications, allowing the supervising psychiatrist to quickly assess a patient's history. Subsequent treatment session notes capture changes in treatment, and the patient's clinical scores on the OASIS and PHQ-3 are graphically depicted over time so that trends can easily be appreciated by ACSs and supervisory personnel. This allows supervisors to monitor patient progress, identify potential problems with the treatment process, and focus discussion on patients who are not progressing optimally. The system allows for “clinical judgment overrides”. When the ACS and psychiatrist agree that the AUDIT or MINI did not accurately capture the diagnosis, the ACS can change the eligibility status and note the reasons for the change. Finally, the system cues the ACS to follow up with patients and record the successful contact or unsuccessful contact attempt. The system also encourages the ACS to record contact with the primary care physician, capturing the “collaborative care”.
Because the system operates in real time, it is especially useful in clinical supervision in that it allows oversight of implementation, standardization of the intervention across sites, and clinical progress. All data communicated between the user's computer and the server are two-way encrypted with the secure sockets layer (SSL) protocol using 128-bit cryptography algorithms. Access to all data on the server is password-protected and user profiles limit access to parts of the database.
We will test the hypothesis that CALM intervention patients will have better outcomes than UC patients. Primary clinical outcomes include symptoms, functioning, and satisfaction with care (Table 3). Outcomes are assessed at baseline and 6, 12, and 18 months. Assessments are nested within patients; patients are nested within providers; providers are nested within clinics; and clinics are nested within sites. Since sites and clinics were purposively sampled they will be treated as fixed effects. An intent-to-treat analysis will be conducted, using time trend models to examine temporal effects. The study has been powered to investigate each of the four diagnostic groups separately for clinical outcomes unique to each group with effect sizes around 0.45 for single disorders. The study has sufficient power for examining the CALM effect on general outcomes, applicable to all disorders with effect sizes of 0.3 overall.
Secondary analyses will use quasi-experimental methods to examine the effect of appropriate anxiety treatment on outcomes. Non-experimental methods will be used to assess the effects of treatment, rather than the effects of the intervention per se. Since it is unlikely that case-mix variables will control fully for unobserved heterogeneity, instrumental variables may be employed. Finally, although CALM is the largest randomized trial of patients with anxiety disorders to date, the study will have limited power to examine health care costs and cost-effectiveness. Therefore, cost analyses will be primarily descriptive.
The CALM study will test an innovative collaborative care model for anxiety disorders. The intervention has been implemented in 12 clinics in four sites in the US. With a population of 1040 subjects, it will be the largest randomized trial to date of persons with anxiety disorders. The CALM intervention represents a step forward in collaborative care models because it is expected to be effective for multiple mental health disorders and to accommodate a range of comorbid conditions. The CALM approach to CBT training and computer-assisted therapy is especially innovative. A number of aspects of the intervention, included to maximize its user-friendliness in the primary care setting, are expected to serve future efforts to disseminate this model of care more widely.
This work was supported by National Institute of Mental Health grants (U01 MH070018 to RAND, U01 MH058915 to UCLA, U01 MH057835 to UCSD, U01 MH057858 to University of Washington, U01 MH070022 to UAMS, MH065324 to Dr. Roy-Byrne and MH64122 to Dr. Stein) and by a VA Health Services Research and Development Service Career Award to Dr. Edlund (RCD 03-036). The authors thank Carrie Edlund, MS, and Christina Reaves-Powell, MPH, for their assistance with manuscript preparation.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.