Hybrid Trial of Brief Cognitive-Behavioral Therapy for Health Anxiety in Primary Care

Seventy-two health anxious patients with a diagnosis of Illness Anxiety Disorder or Somatic Symptom Disorder will be randomly assigned to either receive brief cognitive-behavioral therapy (CBT) administered by a medical assistant in the primary care practice where the patient receives their primary care services (intervention condition) or receive a referral to a mental health provider within the outpatient psychiatry clinic or another community provider of their choice (usual care condition). Randomization will be determined by a random number generation program at the point of enrollment. The randomization scheme will be stratified by primary care site to reduce site effects. The primary care sites from which patients will be recruited include the General Internal Medicine clinic at Dartmouth-Hitchcock Medical Center, Dartmouth-Hitchcock Heater Road clinic, and Dartmouth-Hitchcock Manchester. Recruitment will be evenly distributed across sites. The investigators will over-sample racial and ethnic minority patients at Dartmouth-Hitchcock Manchester due to the higher proportion of minority patients receiving services at that location.

The brief CBT intervention is a 4-session psychotherapy comprised of techniques guided by cognitive-behavioral theory and specifically derived from previously published materials. Each technique has support from prior randomized controlled trials demonstrating their efficacy for reducing symptoms of health anxiety and expert consensus. The techniques employed include motivational interviewing, psychoeducation, cognitive restructuring, and situational and interoceptive exposure. Each session of brief CBT is 45-minutes in duration and is delivered by a medical assistant who is principally located at the patient's primary care practice. All sessions are based on a manual developed by the Principal Investigator that includes essential elements to deliver at each session, scripts to guide delivery of the intervention, and homework assignments for the patient to complete between sessions and after completion of the final session. The Principal Investigator provides clinical supervision of the intervention over the course of the study.

The referral process employed in the usual care condition consists of an electronic referral made by the patient's primary care provider to a mental health provider within the Dartmouth-Hitchcock Health system, or a community provider outside of the system identified by the patient in consultation with the primary care provider. The nature of the intervention provided by the mental health provider is not prescribed or guided by the study protocol.

Examination of distributions of all variables using descriptive analyses will precede inferential statistical analysis. Transformations or recoding will be applied to normalize continuous data when necessary. The success of random assignment will be evaluated by comparing the two study groups on baseline characteristics using chi-square tests for categorical variables and t-tests for continuous covariates. If baseline variables are significantly different between groups and are significantly associated with the outcomes (e.g., medical comorbidity), the investigators will include them as covariates in the analytic models.

The primary outcome of change in health anxiety will be tested using a model that accommodates attrition and is appropriate for correlated data due to clustering (by 3 sites and 6 medical assistants) and repeated measures (baseline, 4 and 12 weeks) on the outcomes. Because there are only three assessment points, time will be treated as discrete and fitted to a covariance pattern model to compare average change between groups over time. Group (CBT vs usual care), time (baseline, 4 and 12 weeks), and the group by time interaction will be specified for the model. The treatment effect of intervention relative to control will be evaluated by testing significance of the group by time interaction term. An unstructured variance-covariance structure will be specified to account for the correlated nature of the data due to repeated measures.

The cluster size at site and provider levels are too small to be treated as random effects; therefore, the analysis will take clustering effects due to site and provider into account using small-sample robust estimate correction options implemented in the Statistical Analysis Software (SAS) generalized linear mixed models procedure (PROC GLIMMIX) procedure. GLIMMIX facilitates conducting mixed-effects and generalized estimating equation (GEE) modeling in the same procedure simultaneously.

The sample size was determined by a combination of recruitment feasibility, study therapist (medical assistant) availability, and power analysis. Given two groups, 3-assessment points, and sample size of 72 (36 per arm) with 15% attrition at 12 months (N = 58), and assuming cross-time correlation of 0.60, alpha of 0.05, and two-tailed tests, the study achieves 80% power to detect a minimum effect size of 0.69, between medium (0.50) and large (0.80) in Cohen's d-metric. If the analysis takes clustering effects due to site and providers into account by adjusting the sample size with an intraclass correlation (ICC) of 0.01, the study achieves 80% power to detect an effect size of 0.72. Statisticians recommend ICC between 0.01 and 0.02 in clinical studies with clustering; the investigators used ICC = 0.01 for this power calculation. A larger sample size is needed to detect effect sizes smaller than 0.69 and 0.72; however, sample size for pilot study should be based on pragmatics of recruitment and the necessities for examining feasibility, and not for purpose of inferential statistical tests.

The investigators will assess acceptability of the intervention and strategy using scores on a self-report measure of treatment acceptability using mean scores above a predetermined cut score denoting moderate acceptability across all participants. The investigators assess study therapist fidelity to the treatment by compiling standardized fidelity ratings for each session, nested within each provider, and converted to a percentage fidelity score.

Finally, the investigators will assess acceptability qualitatively using qualitative interviews with medical assistants, primary care providers, and clinic administrators. Interview data will be transcribed and entered verbatim into a qualitative data management software package for content analysis. Content analysis consists of reviewing the responses of interviewees to identify responses that are coded to correspond to Consolidated Framework for Implementation Research constructs. Coding will be conducted by two independent coders (the Principal Investigator and a research assistant) and overseen by a medical anthropologist.