This is an open-access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in JMIR Mental Health, is properly cited. The complete bibliographic information, a link to the original publication on https://mental.jmir.org/, as well as this copyright and license information must be included.
Anxiety is rising across the United States during the COVID-19 pandemic, and social distancing mandates preclude in-person mental health care. Greater perceived control over anxiety has predicted decreased anxiety pathology, including adaptive responses to uncontrollable stressors. Evidence suggests that no-therapist, single-session interventions can strengthen perceived control over emotions like anxiety; similar programs, if designed for the COVID-19 context, could hold substantial public health value.
Our registered report evaluated a no-therapist, single-session, online intervention targeting perceived control over anxiety in the COVID-19 context against a placebo intervention encouraging handwashing. We tested whether the intervention could (1) decrease generalized anxiety and increase perceived control over anxiety and (2) achieve this without decreasing social-distancing intentions.
We tested these questions using a between-subjects design in a weighted-probability sample of US adults recruited via a closed online platform (ie, Prolific). All outcomes were indexed via online self-report questionnaires.
Of 522 randomized individuals, 500 (95.8%) completed the baseline survey and intervention. Intent-to-treat analyses using all randomized participants (N=522) found no support for therapeutic or iatrogenic effects; effects on generalized anxiety were
Strengths of this study included a large, nationally representative sample and adherence to open science practices. Implications for scalable interventions, including the challenge of targeting perceived control over anxiety, are discussed.
ClinicalTrials.gov NCT04459455; https://clinicaltrials.gov/show/NCT04459455
Social isolation, looming threats of infection, and declining confidence in our abilities to cope have been spurred by the COVID-19 pandemic [
Likewise, levels of anxiety symptoms appear to be rising among US citizens. The COVID-19 pandemic has had wide and enduring negative effects on the mental health of individuals across the life span [
Certainly, not all of these shifts reflect increases in pathological anxiety (versus situation-appropriate worry), but reasons remain for serious clinical concern. Increased time spent worrying about COVID-19 relates to more severe anxiety pathology—both in health care workers [
Perceived control, or one’s subjectively felt ability to control one’s environment and inner experiences, prospectively predicts lower distress during and following numerous uncontrollable stressors, from experiencing sexual assault [
During the present pandemic, one’s perceived control over the circumstances may be (and remain) understandably low: No individual can slow its course single-handedly. In fact, perceived control of one’s environment is largely unrelated to anxiety following circumstantial stressors (eg, undergoing basic military training) [
SSIs consistently demonstrate similar effect sizes to multisession therapies on mental health outcomes [
Online, self-administered SSIs are also more easily, rapidly testable in representative samples than interventions requiring clinician contact (either in-person or remotely). Clinical trials of mental health treatments are generally underpowered [
We therefore evaluated whether an online, self-administered SSI designed to strengthen perceived control over anxiety in the context of the COVID-19 pandemic (Contain COVID Anxiety) increased perceived control over anxiety immediate post-SSI and decreases general anxiety 2 weeks later more than a placebo, handwashing-plan SSI (Remain COVID Free) in a weighted-probability sample of the United States (n=500, See
Our primary hypothesis concerned whether the Contain COVID Anxiety SSI decreased generalized anxiety symptoms 2 weeks later more than the placebo SSI.
Design table.
Question and hypotheses | Sampling plan (eg, power analysis)a | Analysis plana | Interpretation given to different outcomes | ||||
|
|||||||
|
H1: Generalized anxiety decreases more when participants are randomized to the Contain COVID Anxiety SSI than when participants are randomized to the Remain COVID Free placebo SSI. | H1: n=400 for 95% power | H1: Test for assumptions and apply transformations as necessary. Take the 2-week follow-up generalized anxiety mean and enter it as the dependent variable in a linear model with the baseline generalized anxiety mean and treatment condition as predictors. | Support for H1: If the |
|||
|
H0: Generalized anxiety either does not increase more when participants are randomized to the Contain COVID Anxiety SSI than when participants are randomized to the Remain COVID Free placebo SSI, or generalized anxiety increases more when participants are randomized to the Remain COVID Free placebo SSI than when participants are randomized to the Contain COVID Anxiety placebo SSI. | H0: n=150 for 95% power | H0: Test for assumptions and apply transformations as necessary. Take the 2 weeks later generalized anxiety mean and predict it with the baseline generalized anxiety mean. Enter the mean and SD of the standardized residuals from that model for when the condition is Remain COVID Free and the mean and SD of the standardized residuals from that model when the condition is Contain COVID Anxiety into a between-groups equivalence test with equivalence bounds of |
Support for H0: If the |
|||
|
|||||||
|
H1: Perceived control over anxiety increases more when participants are randomized to the Contain COVID Anxiety SSI than when participants are randomized to the Remain COVID Free placebo SSI. | H1: n=350 for 95% power | H1: Test for assumptions and apply transformations as necessary. Take the post-SSI perceived control over anxiety mean and enter it as the dependent variable in a linear model with baseline perceived control over anxiety mean and treatment condition as predictors. | Support for H1: If the |
|||
|
H0: Perceived control over anxiety either does not increase more when participants are randomized to the Contain COVID Anxiety SSI than when participants are randomized to the Remain COVID Free placebo SSI or perceived control over anxiety increases more when participants are randomized to the Remain COVID Free placebo SSI than when participants are randomized to the Contain COVID Anxiety placebo SSI. | H0: n=150 for 95% power | H0: Test for assumptions and apply transformations as necessary. Take the post-SSI perceived control over anxiety mean and predict it with the baseline perceived control over anxiety mean. Enter the mean and SD of the standardized residuals from that model for when the condition is Remain COVID Free and the mean and SD of the standardized residuals from that model when the intervention order is Contain COVID Anxiety into a between groups equivalence test with equivalence bounds of |
Support for H0: If the |
|||
|
|||||||
|
H1: Social distancing intentions do not increase or decrease pre-Contain COVID Anxiety SSIs to immediate post-Contain COVID Anxiety SSI. | H1: n=154 for 95% power | H1: Test for assumptions and apply transformations as necessary. Enter the mean and SD of the social distancing intention composites at baseline and post-Contain COVID Anxiety SSI (only among people who were randomized to the Contain COVID Anxiety) into a paired equivalence test with equivalence bounds of |
Support for H1: The |
|||
|
H0: Social distancing intentions either increase or decrease pre- to immediate post-Contain COVID Anxiety SSI | H0: n=156 for 95% power | H0: Test for assumptions and apply transformations as necessary. Enter the mean and SD of the social distancing intention composites at baseline and post-Contain COVID Anxiety SSI (only among people who were randomized to the Contain COVID Anxiety) into a paired |
Support for H0: The |
aRefer to the R code on the open science framework page for the power analysis and analysis plan [
bSSI: single-session intervention.
We expected a larger decrease in generalized anxiety symptoms from immediately pre-SSI to 2 weeks later when participants were randomized to the Contain COVID Anxiety SSI instead of the placebo SSI. This pattern of results would indicate a larger decrease in generalized anxiety symptoms occurs pre-SSI to 2 weeks later in the Contain COVID Anxiety SSI compared with the placebo SSI.
We also tested whether the active SSI Contain COVID Anxiety can be deployed at scale without reducing intentions to engage in social distancing. Social distancing remains the most potent known nonpharmacological intervention to reduce the spread of the SARS-Cov-2 virus [
We hypothesized that engaging in the Contain COVID Anxiety SSI would have an association with pre- to post-SSI change in social distancing intentions that is statistically equivalent to 0. This pattern of results would indicate completing the intervention is not meaningfully associated with intentions to socially distance.
We were also interested in whether the SSI designed to increase perceived control over anxiety did, in fact, increase perceived control over anxiety immediate post-SSI more than the placebo SSI.
We expected a larger increase immediate pre- to immediate post-SSI in perceived control over anxiety to occur for participants randomized to the Contain COVID Anxiety SSI, relative to those randomized to the placebo SSI. This pattern of results would indicate a larger increase in perceived control over anxiety occurs pre- to post-Contain COVID Anxiety SSI compared with the placebo SSI.
This study tested the efficacy of decreasing generalized anxiety symptoms (Hypothesis 1) and increasing perceived control of anxiety with the Contain COVID Anxiety SSI in a nationally representative sample (Hypotheses 1 and 3) and the safety of testing and making the SSI available at scale (Hypothesis 2). This is a crucial step toward providing open-access, evidence-based resources to help the US population more effectively cope with their anxiety during the COVID-19 pandemic.
The study was approved by the Institutional Review Board (IRB) of Stony Brook University under protocol #2020-00204. As required by US law, a description of this study is available at Clinicaltrials.gov. We have also included a CONSORT checklist in
We recruited a sample of 500 participants representing a weighted-probability sample of the United States through Prolific, an online platform connecting researchers and participants. Prolific is designed specifically for use in the scientific research context, unlike older online crowdsourcing platforms (eg, Mechanical Turk) that were designed for broader use (eg, by marketing and advertising companies to outsource labor) [
This study’s weighted-probability sample was stratified on age, sex, and ethnicity. To help further maximize the generalizability of our findings, there were no prescreening inclusion nor exclusion criteria other than having a Prolific ID, being at least 18 years old (able to provide consent), and residing in the United States. However, study participation required access to, and comfort using, a device connected to the Internet. We also recruited 8 pilot participants from the Prolific platform before recruiting this weighted probability sample to ensure data were being collected properly, and these pilot participants’ data were not used in confirmatory analyses of this study. All pilot data are available on the open science framework page [
We conducted intent-to-treat analyses including all participants who were randomized to a study condition (n=522, see [
We used a between-subjects design; participants were randomized to receive either the active Contain COVID Anxiety SSI (50% allocation) or the placebo Remain COVID Free SSI (50% allocation). The sequence determining randomization of condition was automatically generated using the randomizer within Qualtrics Survey Software (no blocking was used for this randomization), making the randomization process double-blind. To triple-blind our analysis process, the last author (JS) downloaded the data from Qualtrics and recoded the variable indicating to which SSI the participants were randomized before sending the data to the first author (MM) who performed the primary analyses. Therefore, the primary analyses were conducted without the knowledge of which condition is which.
Power analyses were conducted using a “smallest effect size of interest” approach, where we aimed to be powered to detect the smallest effect size corresponding with a subjectively meaningful difference in participants’ experiences [
The entire procedure was conducted online via the Qualtrics Survey Platform, which participants were linked to directly from Prolific. After providing informed consent, participants spent approximately 8 minutes filling out pre-intervention questionnaires including demographics, depression symptoms, generalized anxiety symptoms, self-hatred, access to mental health treatment, and COVID-19–related stressors.
Immediately following answering these questions and immediately prior to the SSI intervention, participants completed the Anxiety Control Questionnaire-Emotion Control (ACQ-EC) scale, the Hand Washing Intentions scale, and several questions about social distancing asked in national surveys to measure their beliefs about the intentions of others to engage in social distancing behaviors like avoiding public spaces and private gatherings.
The participants were then randomized to and spent approximately 8 minutes completing one of the SSIs described in the following sections (which one depended on the number generated by the random sequence from the Qualtrics randomizer described in the previous paragraphs), immediately followed by approximately 2 minutes completing the ACQ-EC scale, the Hand Washing Intentions scale, several questions about social distancing from a standardized Centers for Disease Control and Prevention (CDC) item bank, and the comprehension questions. They were then sent to a Prolific link for compensation (with an incentive of US $2.17 for the 20-minute survey, or US $6.50 per hour) and reminded of the follow-up survey 2 weeks later. Participants also received a reminder through the Prolific platform 2 weeks later to participate in the 2-minute follow-up survey containing a measure of generalized anxiety symptoms, perceived control over anxiety, and an anchor-based question (see [
The Generalized Anxiety Disorder-7 (GAD-7; [
The ACQ-EC [
The following Social Distancing Intentions questions, the secondary outcome of the study, are part of a standardized item bank provided by the CDC [
We used comprehension questions as an initial quality check to ensure participants comprehended the core messages of both SSIs. These questions go beyond traditional attention check items, which can be answered incorrectly even by attentive participants [
This active SSI was developed following current recommendations for evidence-based SSI design to target mental health–related outcomes [
This placebo SSI was developed to mirror the structure of the Contain COVID Anxiety SSI, discuss COVID-19–related content, and do so without as many of the potential active ingredients of effective SSIs. Participants received scientific information about how soap kills the SARS-CoV-2 virus but no neuroscience information related to behaviors or behavior change. Participants were told didactically that there is only one way to wash their hands effectively, by following this 3-step plan: (1) deciding on10 times a day to wash their hands, (2) putting reminders in their calendar or setting alarms on their phone to remind them to wash their hands, and (3) singing happy birthday to their favorite celebrity twice while washing their hands. They then read 3 testimonials from other people who had implemented this plan, but they did not make a plan themselves. They therefore also did not have the opportunity to share their plan to prosocially help others. The entire intervention took approximately 8 minutes and was completed entirely within the Qualtrics survey platform.
We first tested for dropout from the study due to intervention assignment. For example, people could differentially drop out when receiving the active Contain COVID Anxiety. Thus, we tested for differential dropout using a
We then created 2 separate scores for the GAD-7 to reflect baseline and 2 weeks post-SSI scores by taking the mean of the 7 items at each time point (score range at each time point: 0-3). We then created 2 separate scores for the ACQ-EC to reflect baseline and immediate post-SSI by taking the mean of the 4 items at each time point (score range at each time point: 0-5). We also calculated the mean of the 3 Social Distancing Intentions questions (score range at each time point: 1-4) at baseline and immediate post-SSI to calculate composite social distancing intentions scores. Following the creation of these composites, we imputed any missing data using the expectation-maximization and bootstrapping algorithm implemented with Amelia II in R [
Consistent with best practices, we included all predictors from the statistical model (baseline value of imputed outcome, either perceived control over anxiety or social distancing intentions, and intervention order) and all baseline variables expected to be associated with the outcome variable (for generalized anxiety and perceived control over anxiety: Inventory of Depression and Anxiety Symptoms Dysphoria mean score, having received mental health treatment in the past 12 months or not, and self-hate scale mean score; for social distancing intentions: age, gender [male, female, nonbinary], education level, and income level). Imputed data were analyzed using the
We tested whether the Contain COVID Anxiety SSI decreased scores on the GAD-7 immediately pre-SSI to 2 weeks later more than the Remain COVID Free SSI using a linear regression approach. We entered baseline GAD-7 score and condition as predictors of the follow-up GAD-7 mean score. We expected to see a larger decrease in GAD-7 score when the participants were randomized to Contain COVID Anxiety SSI compared with when they were randomized to the placebo Remain COVID Free. This pattern of differences would indicate a decrease in generalized anxiety disorder symptoms to a greater extent due to the Contain COVID Anxiety SSI compared with the Remain COVID Free SSI. We preregistered that a
We tested whether completing both interventions had an effect on social distancing intentions statistically equivalent to 0 using a paired-equivalence test. We entered baseline and postintervention social distancing intentions mean scores and SDs into a paired-equivalence test with equivalence bounds of
We tested whether the Contain COVID Anxiety SSI increased scores on the ACQ-EC more than the Remain COVID Free SSI using a linear regression approach. We entered baseline ACQ-EC scores and condition as predictors of immediate post-SSI ACQ-EC score. We expected to see a larger increase in ACQ-EC score when the participants were randomized to the Contain COVID Anxiety SSI compared with when they were randomized to the placebo Remain COVID Free SSI. This pattern of differences would indicate an increase in perceived control over anxiety to a greater extent due to the Contain COVID Anxiety SSI compared with the Remain COVID Free SSI. We preregistered that a
Of the 529 participants who began the survey, 522 participants were randomized to achieve the weighted-probability sample of 500 (7 participants exited the survey prior to randomization, and 22 participants exited the survey prior to completion of the baseline survey; ie, 94.5% and 95.8% completion rates among individuals who started the baseline survey and among those who were randomized, respectively). All demographics for all randomized participants are reported by treatment condition in
Demographics by treatment condition.
Demographics | Treatment received | |||
|
|
Active Contain COVID Anxiety (n=261) | Placebo Remain COVID Free (n=261) | |
Age (years), mean (SD) | 46.02 (15.65) | 46.19 (15.71) | ||
|
||||
|
American Indian and/or Alaska Native | 0 (0) | 1 (0.4) | |
|
Asian | 19 (7.3) | 16 (6.1) | |
|
African American | 31 (12.0) | 39 (15.0) | |
|
Hispanic or Latino/a | 16 (6.1) | 11 (4.2) | |
|
Native Hawaiian or Pacific Islander | 1 (0.4) | 0 (0) | |
|
White, non-Hispanic | 187 (72.0) | 190 (73.0) | |
|
More than one race | 4 (1.5) | 3 (1.1) | |
|
Other | 3 (1.1) | 1 (0.4) | |
|
||||
|
Agender | 2 (0.8) | 2 (0.8) | |
|
Genderqueer or gender fluid | 3 (1.1) | 0 (0) | |
|
Man | 127 (49.0) | 129 (49.0) | |
|
Trans man | 3 (1.1) | 1 (0.4) | |
|
Woman | 125 (48.0) | 129 (49.0) | |
|
Other | 1 (0.4) | 0 (0) | |
|
||||
|
Asexual | 6 (2.3) | 5 (1.9) | |
|
Bisexual | 20 (7.7) | 18 (6.9) | |
|
Gay | 3 (1.1) | 7 (2.7) | |
|
Heterosexual | 218 (84.0) | 220 (85.0) | |
|
Lesbian | 2 (0.8) | 2 (0.8) | |
|
Pansexual | 3 (1.1) | 6 (2.3) | |
|
Queer | 5 (1.9) | 2 (0.8) | |
|
Questioning or unsure | 2 (0.8) | 0 (0) | |
|
Other | 2 (0.8) | 0 (0) | |
|
Unknown | 0 (0) | 1 (0.4) | |
|
||||
|
Less than high school degree | 1 (0.4) | 2 (0.8) | |
|
High school degree | 27 (10.0) | 29 (11.0) | |
|
Some college, no degree | 74 (28.0) | 67 (26.0) | |
|
Associate degree | 26 (10.0) | 31 (12.0) | |
|
Bachelor’s degree | 77 (30.0) | 99 (38.0) | |
|
Master’s degree | 46 (18.0) | 24 (9.2) | |
|
Professional degree | 4 (1.5) | 4 (1.5) | |
|
Doctorate | 6 (2.3)2 | 5 (1.9) | |
|
||||
|
Less than 10,000 | 17 (6.5) | 14 (5.4) | |
|
10,000-19,999 | 24 (9.2) | 22 (8.4) | |
|
20,000-29,999 | 29 (11.0) | 28 (11.0) | |
|
30,000-39,999 | 23 (8.8) | 24 (9.2) | |
|
40,000-49,999 | 21 (8.0) | 33 (13.0) | |
|
50,000-59,999 | 26 (10.0) | 25 (9.6) | |
|
60,000-69,999 | 15 (5.7) | 24 (9.2) | |
|
70,000-79,999 | 28 (11.0) | 20 (7.7) | |
|
80,000-89,999 | 15 (5.7) | 10 (3.8) | |
|
90,000-99,999 | 15 (5.7) | 12 (4.6) | |
|
100,000-149,999 | 27 (10.0) | 30 (11.0) | |
|
150,000 or more | 21 (8.0) | 19 (7.3) | |
|
||||
|
No current relationship | 101 (39.0) | 118 (45.0) | |
|
Relationship, not living together | 25 (9.6) | 20 (7.7) | |
|
Relationship, living together | 24 (9.2) | 22 (8.4) | |
|
Engaged | 3 (1.1) | 3 (1.1) | |
|
Married | 108 (41.0) | 98 (38.0) | |
Has children, n (%) | 117 (45.0) | 127 (49.0) | ||
Health insurance covers mental health, n (%) | 195 (75.0) | 196 (75.0) | ||
Received psychotherapy in the past year, n (%) | 56 (21.0) | 54 (21.0) | ||
Received medication for mental health in the past year, n (%) | 56 (21.0) | 54 (21.0) | ||
Perceived need for mental health treatment in the past year, n (%) | 87 (33.0) | 89 (34.0) | ||
Baseline IDASa-Dysphoria (1-5), mean (SD) | 2.00 (0.90) | 2.09 (0.92) | ||
Baseline GAD-7b (0-3), mean (SD) | 0.75 (0.74) | 0.77 (0.75) | ||
Baseline self-hate (1-7), mean (SD) | 2.07 (1.65) | 2.20 (1.59) | ||
Baseline perceived control over anxiety (0-5), mean (SD) | 2.88 (1.34) | 2.85 (1.29) | ||
Baseline Hand Washing Intentions (1-7), mean (SD) | 5.08 (1.59) | 5.01 (1.61) | ||
Baseline social distancing intentions of others (1-4), mean (SD) | 2.27 (1.12) | 2.32 (1.12) |
aIDAS: Inventory of Depression and Anxiety Symptoms.
bGAD-7: Generalized Anxiety Disorder-7.
There was no evidence participants were significantly more likely to drop out of either condition at the 2-week follow-up (25/261, 9.6% dropped out from the Remain COVID Free SSI, and 18/261, 6.9% dropped out from the Contain COVID Anxiety SSI;
During our piloting of the Prolific platform (as outlined in our preregistered message), we noticed a substantial portion of participants were not answering the comprehension check questions correctly despite providing face-valid qualitative and quantitative data. We updated our comprehension check questions to attempt to align them more with completing the intervention with fidelity. However, among all participants who were randomized to the Contain COVID Anxiety SSI and answered a comprehension check question, 52.3% (126/241) answered both comprehension questions correctly. To examine whether this phenomenon was a function of the questions or lack of fidelity to the intervention, we developed a systematic qualitative coding system focused on fidelity for each qualitative response in the Contain COVID Anxiety SSI group. To be coded as having a high-fidelity qualitative response, the participant had to respond not only to the prompt with related content (a more general comprehension check) but also to the prompt as instructed (eg, a response enumerating concrete coping strategies to a prompt instructing participants to validate their own anxiety would be marked as a low fidelity response; see the publicly available code for the full qualitative coding system for fidelity check [
We double-coded a random 20% of intervention responses (48 participants with 6 responses each, effective n=288) and found 87.13% average fidelity across these participants’ responses. Further, answering both comprehension check questions correctly shared only 0.01% of the variance with each participant’s fidelity score across their qualitative responses. We therefore determined that the comprehension check questions were poor indicators of completing the intervention with fidelity and chose to proceed with our planned analyses.
In full intent-to-treat analyses with all participants who were randomized (n=522), we did not find support for the alternative hypothesis (
To make it possible to generate fully invertible matrices necessary to produce imputations, participant gender was dropped from the imputation model. In this case, the alternative hypothesis was operationalized as a change in social distancing intentions of others pre- to immediate post-Contain COVID Anxiety being statistically equivalent within a range of
In full intent-to-treat analyses with all participants who were randomized (n=522), we did not find support for the alternative hypothesis (
Compared with a placebo control, a self-guided SSI for US adults did not improve short-term generalized anxiety or perceived control over anxiety during the COVID-19 pandemic. This high-powered randomized controlled trial (RCT), which used a nationally representative US sample, also demonstrated that this intervention did not worsen short-term generalized anxiety or perceived control. There was also statistically equivalent to zero iatrogenic movement within the intervention condition of beliefs in others’ willingness to social distance.
Interest in the use of brief, e-mental health interventions has increased substantially during the COVID-19 pandemic across the general adult population [
Accordingly, these results are the first to suggest that perceived control over anxiety and generalized anxiety symptoms may in fact be difficult to move in general adult samples via self-guided SSIs, at least in this nationally representative sample. Within-group effect sizes for perceived control over anxiety was negligible in both the active and placebo conditions, in contrast to within-group SSI effects seen in trials targeting adolescents. Further, nonzero increases were observed in generalized anxiety symptoms in both the active and placebo conditions over 2 weeks. Therefore, it is not the case that participants benefited from
This design did not contain a wait list control condition, and we cannot explicitly rule out that receiving either light-touch intervention would have resulted in a smaller increase in generalized anxiety disorder symptoms compared with receiving nothing. This pattern of within-group effect sizes (ie, increasing generalized anxiety symptoms over time in both conditions) is consistent only with potentially preventative, as opposed to therapeutic, on average effects compared with “no treatment” control. Although we found no evidence of iatrogenic movement on social distancing intentions of others within our SSI, the lack of iatrogenic effects in other e-mental health interventions cannot be guaranteed without testing those outcomes directly. E-mental health applications hold promise in increasing mental health treatment access, [
We would like to propose 2 complementary paths toward building and understanding the impacts of effective SSIs for anxiety in adults, based on the results of this trial, which may generalize to evaluations of other light-touch interventions as well. First, given substantial heterogeneity in individual-level responses to any mental health intervention (including the SSI tested here), we recommend that researchers and program developers collect data necessary to build predictive models of individual-level response to SSIs. Predictive models require much larger sample sizes than typical clinical trials collect to identify subgroups of best responders. For example, recent simulation studies demonstrated that clinical trials may need as many as 500 participants per treatment arm to recover reliable predictions about who would benefit most from which treatment (ie, questions of moderation effects; [
Second, we recommend the systematic incorporation of qualitative and user-experience data into trials of self-guided SSIs. It has been posited that SSIs targeting adolescent mental health problems may show acceptability and efficacy, at least in part, because they do not “feel” like interventions to youth—that is, they are designed to be nonstigmatizing to users [
There are certainly limitations to what this study can conclude. First, this study was conducted during the COVID-19 pandemic, and it is unclear whether the nonzero increases in generalized anxiety within both groups reflected the many structural challenges of pandemic conditions (which a self-guided SSI cannot change) or would have occurred regardless. Examination of within-group effect sizes in self-guided SSI trials conducted after the COVID-19 pandemic ends should examine whether negligible to slightly increasing within-group effect sizes persist for clinical anxiety in unselected adult samples. Other work suggests that certain outcomes, such as parenting behaviors and distress tolerance, may be modifiable via self-guided SSIs in high-symptom adults even during the pandemic [
Compared with a placebo control, an 8-minute, self-guided SSI for US adults did not improve short-term generalized anxiety nor perceived control over anxiety during the COVID-19 pandemic. Additionally, neither condition yielded any iatrogenic movement in a key public health behavior (assumed social distancing intentions of others). Our rigorous methods and well-powered sample bolster confidence in these results, which carry direct implications for future research on self-guided SSIs for mental health problems—both for anxiety in adults and more broadly. As with so many interventions targeting complex, individual-level problems, key questions for SSI research remain: “which intervention, for whom, and under what circumstances?” [
Full survey and interventions.
CONSORT-eHEALTH checklist (V 1.6.1).
COVID Anxiety SSI RR preprocessing and analyses.
Anxiety Control Questionnaire-Emotion Control
Centers for Disease Control and Prevention
Generalized Anxiety Disorder-7
institutional review board
randomized controlled trial
single-session intervention
This work was supported by research funding from the Department of Psychology at Stony Brook University to JS. We thank Ms. Sharon Chen, Mr. Akash Shroff, and Ms. Julia Fassler for piloting and providing key feedback on survey and intervention materials evaluated in this study.
Data from this study are publicly available [
MM, MD, JS, and JS contributed to study conceptualization. MM, JS, MD, and JS contributed to the intervention design. MM wrote the original draft of the manuscript. JS, MD, JS, IA, CB, and JS contributed to manuscript editing and revisions. MD wrote the original draft of the primary analysis code. JS wrote the original draft of the power analysis code. MM edited the analysis code/power analysis code. MM, MD, IA, and JS will be responsible for data curation, investigation, and formal analysis. JS contributed to funding acquisition and supervision.
The funders have/had no role in study design, data collection and analysis, decision to publish or preparation of the manuscript. Unrelated to the present study, JS presently receives grant and research support from the National Institutes of Health (DP5OD28123), the Klingenstein Third Generation Foundation, the American Psychological Foundation, and Limbix Inc; all grants support research evaluating brief or single-session mental health interventions. MD receives grant support from a Stony Brook University Graduate Research Fellowship, and MD and JS receive research funding from the Psi Chi Honor Society for research on single-session interventions. JS, MD, and MM are under contract with New Harbinger Publications to co-author a therapeutic workbook for adolescents. JS is under contract with Oxford University Press to co-edit a book on low-intensity mental health interventions for youth. CB reports receiving grant support from the National Institute of Health and compensation for his work as journal editor for Sage Publications Inc. The authors report no other financial or nonfinancial conflicts.