Treating Young Adult Generalized Anxiety Disorder With Text-Message Delivered Cognitive Behavioral Therapy

All participants are informed in the consent procedures (consent form) that we are testing the CBT-txt-Anxiety intervention, that it is based on clinical procedures that have been shown to be effective in in-person and web-based counseling modes but has not yet been proven to be effective in text-mode, and that participating in this study does not restrict them from seeking additional forms of counseling. Aim 1: Evaluate the feasibility of the study's recruitment procedures and CBTtxt-Anxiety's acceptability and participant satisfaction. Hypothesis 1A: Study protocols will meet or exceed benchmark criteria for protocol processes (recruitment rates, time, technical issues). Hypothesis 1B: Participants' reports of acceptance and satisfaction (assessment burden, timing of texts, content) and engagement (completion rate, response time, utilization) will meet or exceed benchmark criteria. Aim 2: Test the efficacy of CBT-txt-Anxiety with young adults to reduce anxiety symptoms. Hypothesis 2: Participants receiving CBT-txt-Anxiety will have fewer anxiety symptoms at 2- months post-intervention compared to waitlist controls. Study Population \& Inclusion/Exclusion Criteria: We will use social media to recruit 100 young adults (ages 18-25) who meet criteria for elevated anxiety symptoms. National epidemiological data estimates that one in five (19.5%) of young adults experienced generalized anxiety disorder in the past year compared to one in six (15.6%) adults aged 18 or over (Terlizzi \& Villarroel et al., 2020) and of those, approximately half (52%) also report elevated depressive symptoms (Kessler et al., 2015). Based on our past success identifying and recruiting participants into our research studies, we conservatively estimate that we will be able to recruit and enroll 10 participants into the study. We estimate that we may consent up to 1500 participants, in order to enroll 100 - due to screening out those who do not meet eligibility criteria as well as those likely to be scammers or 'bots'. Study Procedures: All procedures - interactions/interventions following recruitment and screening - have been designed to be virtual/on-line. 3A. Baseline survey - All participants who consent to participate and screen into the study, will be directed to the baseline survey that will be conducted online via Qualtrics (Qualtrics,2020). The survey is estimated to take approximately 30 minutes to complete. 3B. Randomization - Once Qualtrics determines the participant has completed the baseline survey, Qualtrics' automated randomization procedures will randomly assign participants to either CBT-txt-Anxiety or waitlist control. We will use a randomized 10-block design, blocking on study arm and gender, and an allocation ratio of 1:1 in order to have a balanced design. Once randomized participants are notified after their survey is complete with an automated message from Qualtrics. 3C. CBT-txt-Anxiety Intervention - Participants assigned to the CBT-txt- Anxiety intervention condition will receive 32 text conversations over a 60-day period. Participants indicate what time of day they would like to receive the intervention texts. The texts delivered in the CBT-txt-Anxiety intervention focus on empowering participants to understand how thoughts, activities, and other people affect their mood and anxiety. Texts are individualized, based on data provided by participants in the baseline survey. Follow-up surveys - All participants will be asked to complete an online survey at 1-, 2-, and 3-months post-baseline. Participants will be notified by text and directed to the online Qualtrics survey. 3E. Ecological Momentary Assessment - All participants in the treatment group only will be asked to complete 1 ecological momentary assessment via text message every day for 60 days. These assessments will be collected immediately prior to the text counseling conversation on intervention days. Preliminary Analyses - We will begin by plotting outcome distributions and making normality transformations where needed, visualizing change in our outcomes over time via longitudinal plots and taking stock of missing data. 4B. Hypothesis Testing Aim 1: Evaluate the feasibility of the study protocol's processes, resources, and management, and the treatment's acceptability. Hypothesis 1A: Study protocols will meet or exceed benchmark criteria for protocol processes (recruitment and enrollment rates, time), resources and management (recruitment cost, staff time, technical issues). Hypothesis 1B: Participants' reports of acceptance and satisfaction (assessment burden, timing of texts, content) and engagement (completion rate, response time, utilization) will meet or exceed benchmark criteria. Analytic Approach 1: We will use descriptive statistics for rates of recruitment, enrollment, acceptability, satisfaction, and engagement in our sample and report whether they meet or exceed our hypothesized levels. Ttests will be used to compare sample differences in time to enrollment, acceptance, and satisfaction scores; chi-square tests will be used to compare engagement rates with our previous pilot studies. Aim 2: Test the efficacy of CBT-txt-Anxiety with young adults to reduce anxiety symptoms. Hypothesis 2: Participants receiving CBT-txt-Anxiety will have fewer anxiety symptoms compared to waitlist controls Analytic Approach to Aim 2: Tests for Hypothesis 2 will be performed using a full information maximum likelihood structural equation modeling (SEM) framework, which accounts for missing data, boosts power and is consistent with intent-to-treat models. We will begin by running unconditional latent growth models to estimate change in outcomes (anxiety symptoms) over the 4 data collection periods (baseline, 1 month, 2 months, and 3 months). For each outcome, the unconditional models will estimate an intercept (baseline level of the outcome), a linear slope (change in the outcome with each passing month), and random effects for each, whose variances and covariance account for individual differences in the growth trajectories (e.g., higher versus lower baseline levels; faster versus slower rates of monthly change). If the linear slope is significant in the unconditional model, we will investigate polynomial functions to capture potential nonlinear change functions, starting with quadratic, and moving to cubic only if the quadratic function is significant. Once the functional form of change is identified in the unconditional model, we will estimate differences in the growth parameters by experimental condition (CBT-txt-Anxiety). Differences by treatment group will be estimated for the baseline level of outcome variables (anxiety symptoms) and for the monthly rate of change over time on those variables. Significance tests for these associations will inform whether young adults randomized to receive CBT-txt-Anxiety show different baseline levels and faster rates of decline in outcomes (anxiety symptoms) compared to young adults randomized to waitlist control, with the latter test providing evidence of a statistically significant treatment effect. Power Analysis - Our power analysis is based upon 100 participants, with 50 allocated to the treatment condition and 50 allocated to the wait-list control condition. With 100 participants, medium effect sizes (standardized treatment effects of .3 - .4 SD units) are needed to achieve 80% power. We conducted power analyses by conducting Montecarlo simulations of growth curve analysis in Mplus. We generated 4 waves of data (baseline, 1 month, and 2 months, and 3 months) with steadily increasing probability of missing so that 10% of the sample were missing at the 2-month follow-up. Following Muthén and Curran (1997), we specified the following modeling conditions. We specified the slope factor such that (a) the increase from the baseline to the 2-month follow-up corresponded to 1 SD increase in the outcome, and (b) the variance of the slope factor was set to 20% of the variance of the intercept factor. We assumed that due to randomization, the treatment and control groups would not differ at the initial time point, thus the treatment effect on the intercept was set to 0. We set the treatment effect on the slope to be equal to the difference, in SD units of the outcome, between the treatment and control groups at the 3- month follow-up. Simulations revealed that, under these conditions and with an initial sample size of 100, a treatment effect of standardized =0.31 was sufficient to produce 80% power.