Brief Imaginal Exposure Exercises for Social Anxiety Disorder: A Randomized Controlled Trial of a Self-help Momentary Intervention App

Abstract

Introduction:

Emerging technological interventions for psychological disorders are being developed continually. Offering imaginal exposure exercises as a self-help intervention presents the opportunity to acquire foundational skills to address social anxiety. The current study evaluates the feasibility and effectiveness of a novel smartphone application for social interaction anxiety.

Methods:

Participants ($n=82$) were adults meeting criteria for social anxiety disorder. They were randomly assigned to imaginal exposure (IE; $n=39$) or self-monitoring ($n=43$) delivered multiple times daily via a smartphone application for a one-week trial. It was expected that participants using the IE exercises would demonstrate significantly greater declines in social anxiety in addition to increases in self-efficacy and that compliance would serve as a predictor of outcome. Mixed-effects models were utilized.

Results:

Participants using IE (vs. self-monitoring) evidenced significantly greater reductions in social anxiety from pre- to post-treatment and at 1-month follow-up. Similarly, IE (vs. self-monitoring) led to significantly greater increases in self-efficacy from pre- to post-treatment and 1-month follow-up. Further, more completed IE exercises predicted significantly greater changes in social anxiety and self-efficacy at subsequent timepoints compared with self-monitoring.

Conclusions:

Findings suggest that a brief IE self-help intervention was effective in targeting social interaction anxiety.

Social anxiety disorder (SAD) is highly prevalent among university undergraduates (Auerbach et al., 2016; Eisenberg et al., 2013). In individuals with SAD, social and performance situations may be avoided or experienced with severe distress (Harris et al., 2002). SAD has a 12-month prevalence of 7% in the United States and has displayed associations with elevated rates of school dropout, as well as decreased well-being, employment, productivity, and quality of life (American Psychiatric Association, 2013). Only about 20 to 40% of undergraduate students with SAD reported receiving mental health support (Eisenberg et al., 2011), leading to an estimated 4 million university undergraduates with untreated mental illnesses (Kazdin, 2017). Thus, finding ways to disseminate helpful treatments to this group is important.

Exposure has proven to be an efficacious and primary method of intervention for individuals with social anxiety disorder (SAD; Kaczkurkin & Foa, 2022; Mayo-Wilson et al., 2014; Ponniah & Hollon, 2009). Typically, such exposure has been completed in-vivo. Efforts to supplement in vivo exposure treatment with social skills training have displayed further clinical and skills-based benefits (Beidel et al., 2014). However, difficulties exist with in vivo exposure, including rarity of exposure opportunities, lack of access to anxiety provoking stimuli or situations, lack of practitioners trained in and willing to provide exposure, and lack of patient compliance with the exercises (Pittig et al., 2019). Virtual reality exposure (VRE) has been employed to offset some of these issues (Chesham et al., 2018; Kampmann et al., 2016; Zainal et al., 2021), However, VRE is not particularly accessible to most people. Also, standardized videos used in VRE and other types of video exposure do not allow for individualizing exposure hierarchies.

One solution to providing accessible individualized exposure is via imaginal exposure (IE). Drawing on the conceptual imagery model proposed by Lang (1977), creating successful IE has challenges. Specifically, important imagery skills in IE include the skill of the therapist leading the client through the exposure exercise, as well as skills of the individual engaged in the exposure to create imagery and produce an emotional response. Producing an emotional response from IE could also require presence. Presence is the ability to feel completely immersed and involved in what is being imagined as though the imagined events were actually happening. Although presence has been primarily studied in VRE (Schubert et al., 2001), it may be particularly relevant for IE wherein there are no tangible stimuli. To be efficacious, IE rests upon the efforts of the individual to do everything in imagery. Also, whereas presence has historically been considered to be a static phenomenon, a recent study found that it can increase across exposure sessions (Zainal et al., 2021). Thus, it may be possible to get better at it with practice. When all imagery skills are facilitated, the imagined situations in IE create a physiological response similar to in vivo exposure (Costanzo et al., 2014), while increasing engagement (Dinh et al., 1999). These physiological responses (e.g., electrodermal activity and heart rate variability) have been shown to be important activation indicators for individuals with SAD during exposure (Christian et al., 2022) and may be indicators of adequate activation during IE.

A means to increase the dissemination and ecological validity of IE would be to deliver it via a smartphone application. Many studies have been conducted delivering smartphone intervention packages to college students (Fitzsimmons-Craft et al., 2021; Newman et al., 2021a; Newman et al., 2021b). However, literature is lacking with respect to smartphone delivered IE as a stand-alone intervention in those with social anxiety disorder. Unfortunately, a package of multiple techniques does not tell us which treatment components led to observed improvements. In addition, it could be the case that one simple technique is sufficient to treat SAD. Further, focusing on only one technique provides more time to practice it, an important aspect of exposure treatment (Hope et al., 1993). If the smartphone delivery is done in a way to allow individuals to personalize their fear hierarchies it could also take full advantage of the benefits of IE as a self-help option.

Treatment focusing on one technique may also allows for greater efficiency if it is delivered across a short time period. Brief psychological interventions have been defined as treatments provided over a period of 14 days or less (Schumer et al., 2018). Though not administering IE, 10-day (LaFreniere & Newman, 2016) and 14-day (Zainal & Newman, 2023) ecological momentary interventions (EMI) that entailed one technique led to change compared to a self-monitoring control condition for generalized anxiety disorder. Also, fewer IE treatment sessions for post-traumatic stress disorder (PTSD) were not significantly less effective than full length IE (van Minnen & Foa, 2006). In addition, a one-session trial of IE for social anxiety led to changes in core beliefs (Romano et al., 2020). Pairing brevity and repetition could have beneficial impacts for social anxiety exposure and could also leverage the convenience of advanced technology. Also, using briefer smartphone interventions has been proposed as a way to increase usage and engagement (Fleming et al., 2018; Simblett et al., 2018).

Therefore, significant gaps in understanding treatment through IE remain within the SAD literature. Whereas VRE and in vivo exposure are commonly used, IE is an underutilized technique. Dissemination of IE could also be maximized by delivery via a self-help smartphone application. Additionally, IE has not typically been studied in isolation from other treatment techniques. Further, no studies exist involving a brief self-help smartphone app to deliver IE for SAD. Addressing each of these gaps may increase accessibility to exercises typically reserved for delivery during the course of traditional CBT.

The current study examined the efficacy and effectiveness of a 7-day self-help IE Intervention (vs. active control) on individuals with social anxiety. For this study, a neutral task involving self-monitoring was the active control condition, which did not lead individuals to address aspects of their social anxiety (i.e., avoidance) or their overall self-efficacy. It was expected that compared to the active control condition, anxious individuals randomly assigned to undergo the 7-day IE exercises would experience significantly greater reductions in social anxiety. Related to this hypothesis, IE (vs. self-monitoring) was expected to produce greater increases in self-efficacy. Also, individuals using IE who had higher exercise usage were expected to show greater changes in social anxiety and self-efficacy. Finally, we anticipated that presence and mental imagery would become stronger across IE exercises, viewing mental imagery as an adaptable skill.

Method

Participants

Participants were 82 persons meeting criteria for social anxiety disorder (SAD) who were randomized to either 7-day IE exercises ($n=39$) or a self-monitoring active control ($n=43$). Participants were recruited from an undergraduate psychology subject pool mass testing, as well as from advertisements. To be eligible, participants had to be 18 years of age or older, be fluent in English, own an iPhone, and meet DSM-5 criteria (American Psychiatric Association, 2013) for social anxiety disorder (SAD). Diagnostic comorbidity with other psychological difficulties included depressive symptoms (17%), major depressive disorder (5%), and other anxiety disorders (23%) in the sample. Potential participants were excluded if they endorsed mania, psychosis, suicidality, alcohol or substance use disorders, or any medical or organic disorder that hindered their participation in the study. Additionally, this study excluded individuals currently in psychological or psychiatric treatment (including prescribed psychotropic medications) for anxiety or any other mental health issues to isolate and evaluate IE as a stand-alone technique. Of the 82 participants analyzed, 5 dropped out of the study prematurely (2 in IE and 3 in control), though they were retained in the analyses in line with the intent-to-treat principle (see Figure 1 for Consolidated Standards of Reporting Trials (CONSORT) Flowchart). Average age of participants was 19.40 years (SD = .64). There were 62 White participants, 4 African American participants, 2 Hispanic, 8 Asian, and 6 Middle Eastern participants. There were 38 males and 44 females. Please see Table 1 for participant demographics distributed by condition.

Figure 1.

CONSORT Flowchart Depicting Participant Recruitment, Selection, and Randomization, Post-assessment and Follow-up Completion, and Data Analysis

Note: CONSORT = Consolidated Standards of Reporting Trials; IE = imaginal exposure.

Table 1.

Demographic Characteristics by Condition

	SM		IE
	$n$	%	$n$	%
Race
White	32	74	30	77
African American	2	5	2	5
Hispanic	1	2	1	3
Asian	5	12	3	8
Middle Eastern	3	7	3	8
Gender
Male	19	44	19	49
Female	24	56	20	51
	M	SD	M	SD
Age	19.46	0.62	19.32	0.67

Note: $N=82$. SM = self-monitoring control. IE = imaginal exposure therapy.

Measures

The Social Interaction Anxiety Scale (SIAS; Mattick & Clarke, 1998) is a 20-item measure of anxiety in interpersonal interactions. Following psychometric evidence suggesting the removal of the three negatively keyed items (Rodebaugh et al., 2011), only the 17 positively keyed items were used for study analyses. Items were rated on a 5-point Likert scale ranging from 0 (not at all characteristic or true of me) to 4 (extremely characteristic or true of me). The SIAS possesses very strong retest reliability of $r=.92$ (Mattick & Clarke, 1998), strong internal consistency in the current study ($\alpha \text{s}=.79,.83,.81$), and good convergent and discriminant validity in diverse samples (Heimberg et al., 1992).

The Social Phobia Diagnostic Questionnaire (SPDQ; Newman et al., 2003) is a 29-item measure of of social anxiety related to fear and avoidance. The measure has adequate retest reliability ($\kappa =.63$), good convergent validity and discriminant validity, and strong internal consistency in the current study ($\alpha \text{s}=.92,.93,.90$). Additionally, it has great sensitivity (82%) and specificity (85%; Newman et al., 2003).

Diagnostic Interview for DSM-5 Anxiety, Mood, and Obsessive-Compulsive and Related Disorders (DIAMOND; Tolin et al., 2018). The DIAMOND is a semi-structured interview based on the DSM-5. Interrater reliability ranged from $\kappa =.62$ to $\kappa =1.00$ for all diagnoses and the retest reliability was generally high ($\kappa =.59\sim 1.00$). Convergent validity was verified for most of the diagnoses by higher scores on corresponding self-report measures (Tolin et al., 2018). Undergraduate research assistants highly trained by a clinical psychology doctoral student conducted the DIAMOND interviews and were supervised by a doctoral-level licensed clinical psychologist. Interviewers were uninformed of treatment allocation. Each interview was recorded and rated by a second blind rater. In the current study, interrater agreement was excellent (Cohen’s $\kappa =.82,95\text{\%}$ CI [.72,.93]) for the SAD diagnosis.

Avoidance.

After each exercise, participants self-reported the number of avoided situations occurring since the prior completed exercise. Participants were specifically asked “How many social situations/interactions did you engage in since your last response?” and “How many social situations/interactions did you avoid since your last response?”

The General Self-Efficacy Scale (Schwarzer & Jerusalem, 1995) is a 10-item measure of ability to persist and adapt in daily life. Participants rated items on a 4-point Likert scale from 1 (Not at all true) to 4 (Exactly true). The measure has displayed adequate retest reliability ($r=.75$) and possesses high internal consistency ranging from .76 to .90 across 23 different samples (Schwarzer et al., 1999). The internal consistency was also high within the current study ($\alpha s=.84,.82,.82$).

The Igroup Presence Questionnaire (IPQ; Schubert et al., 2001) is a 14-item scale where participants endorse the extent to which the exercises they engaged with felt realistic following the exercise completion. Each response was provided on a 7-point Likert scale. High internal consistency was displayed in the current study (average of $\alpha =.87$) and construct validity has been demonstrated (Vasconcelos-Raposo et al., 2016).

The Questionnaire Upon Mental Imagery (Sheehan, 1967) is a 35-item measure using a 7-point Likert scale which asks participants to rate how vividly they can see imagery from 1 (perfectly clear and vivid) to 7 (no image present at all). This measure displays good internal consistency (Evans & Kamemoto, 1973) which was displayed in the current study (average of $\alpha =.82$) and adequate convergent validity (Sheehan, 1967).

The Post-exercise Imagery Questions (Borkovec & Sides, 1979) were 2 additional imagery vividness items added, also following the completion of each IE exercise. These two questions asked participants to “Please rate how vivid the imagery was during the exercise” and to “Please rate how much fear the imagery created during the exercise.”

Compliance.

Exposure compliance was calculated by dividing the total number of completed exercises by the requested number of prompts completed, per condition. Those in IE were asked to complete exercises at least thrice daily for a total of 21 prompts for compliance calculation. Those in self-monitoring were asked to complete exercises 8 times daily for a total 56 prompts for compliance calculation. This difference in daily exercise completion requests was due to the approximate length of each exercise (self-monitoring = 5 minutes; IE = 13 minutes), as well as a quiet, private place necessary for IE exercise completion.

Procedure

The study was approved by an Institutional Review Board at the participating university. All participants consented to the study. SAD diagnosis was screened using the SPDQ (Newman et al., 2003). Participants invited from screening provided verbal consent during an initial session, followed by the DIAMOND (Tolin et al., 2018) to confirm the diagnosis. If diagnosis was confirmed by the DIAMOND, eligible participants completed baseline symptom self-report measures which were counter-balanced to rule out order effects. See Figure 2 for an overview of participant flow throughout the study.

Figure 2.

Effect of Therapy Condition on Outcome

Note. Outcome (social anxiety or self-efficacy) composite scores are plotted on the y-axis. SM = Self-monitoring; IE = imaginal exposure. FU = follow-up.

Those invited to participate provided their written consent. A randomizer was used to determine treatment allocation for each participant (either IE or active control based), which is an established method in clinical trials (Padhye et al., 2009). Experimenters were blinded to treatment allocation. All participants underwent their assigned exercise condition for 7 days. The app prompted them 8 times daily to complete their assigned exercise. Participants in the IE condition completed questions immediately following each exercise to capture presence and imagery data. To assess intervention effects and maintenance of study gains, participants completed questionnaires at pre-treatment, post-treatment, and 1-month follow-up. The ImExpsoure application was removed from participants phones following their participation in the study at the post-treatment session to observe lasting effects of the intervention at subsequent follow-up. All study sessions were conducted remotely, as the trial was conducted during the active phase of the COVID-19 pandemic with inclusive dates of participation occuring between April 2020 and March 2022.

IE Therapy Condition.

The IE therapy condition provided participants with a therapeutic rationale that repeated practice of IE exercises would help them learn to tolerate anxious distress and overcome avoidance behaviors toward feared situations and increase self-confidence. IE exercises were completed using the ImExposure smartphone application on a participant’s personal smartphone. The application was available in the Apple App Store but was locked with a researcher code. Core features of the application involved engaging psychoeducation videos, a personalized fear hierarchy, imaginal exposure exercise audio, and post-exercise questions. Psychoeducation covered topics of anxiety, avoidance, exposure, imaginal exposure, fear hierarchies, and translating imaginal exposure into in vivo exposure. The personalized fear hierarchy provided participants the opportunity to order the social interaction situations they wanted to address and provide SUDS (Wolpe & Lazarus, 1967) ratings for each. The exercise audio provided a 13-minute imaginal exposure audio exercise which guided participants through a selected exposure from their hierarchy. Exercises were directed toward exposure to a variety of social speaking situations (e.g., talking with peers, talking with a teacher or boss, participating in class, a job interview, giving a public speech, interactions at a party or bar, attending a formal event, asking someone out on a date, etc.). Similar to VRE, the feared stimulus or situation was not physically present during the exercises. However, in IE without VR individuals created the mental imagery themselves in accordance with a series of guiding phrases, rather than viewing the imagery through a technological device. The imaginal exposure prompts were general enough to allow for applicability to any of their listed social speaking situations. Following the completion of each exercise, participants completed reflective pre, peak, and post SUDS ratings related to the imagined situation, Post-exercise Imagery Questions, the Igroup Presence Questionnaire, and follow-up questions about the number of situations avoided and any in vivo exposure they had engaged in since their last completed exercise. After each completed IE exercise, they were also asked to identify a situation from their fear hierarchy for which they would consider completing via an in vivo exposure before the next prompt. Participants in the IE condition were requested to complete imaginal exposure exercises at least three times daily and as many in vivo exposure exercises as they were willing to complete without specified requirements. If prompts were missed, participants were still able to complete an exercise whenever they opened the application at an available time. An excerpt from the IE exercise instructions (see Appendix A) and screenshot of the participant view of the smartphone application during an exercise (see Figure 3) are provided in the Supplemental Materials.

Self-Monitoring Control.

The self-monitoring condition provided participants with a therapeutic rationale that learning and implementing self-monitoring skills would help them manage and lower their anxiety by being able to better recognize anxious symptoms and negative emotions and adjust their avoidance behaviors toward fear and increase their self-confidence in response. This condition asked participants to observe and log their avoidance behaviors related to social situations via a survey link that was emailed to them at the time of each prompt. Participants were to log the date at the time of entry, the number of social speaking opportunities encountered since the last prompt, the number of avoidance behaviors since the last prompt, their average and peak levels of anxiety through subjective units of distress (SUDS) ratings, and the actual speaking situations themseleves. Each self-monitoring exercise prompt took approximately 5 minutes to complete. An excerpt from the self-montoring exercise instructions (see Appendix B) and screenshot of participants’ view of the self-monitoring exercise (see Figure 4) are provided in the Supplemental Materials.

Sample Size Estimation

We aimed to recruit at least 30 participants per treatment condition. A Monte Carlo simulation study with 500 simulations in the $R$ package SIMR (Green & MacLeod, 2016) suggested that the achieved sample of 82 participants provided 81.4% power to detect a small- to medium-sized (Cohen’s $d=.35$) difference in symptom change from pretreatment to posttreatment across the treatment and control groups.

Data Analyses

Preliminary analyses were conducted to examine if there were any differences across the treatment groups in terms of demographic characteristics. Specifically, we used a Fisher’s exact test to compare treatment groups on race, a logistic regression to compare treatment groups on gender, and analysis of variance (ANOVA) to compare groups on age.

Each of the outcome measures collected at pre-treatment, post-treatment, and 1-month follow-up were standardized. The SIAS and SPDQ were standardized and combined to create a social anxiety outcome composite score which has been shown to increase reliability and validity of assessment (Nunnally & Bernstein, 1994) and lower experiment-wise Type I error rate relative to separate analyses for single measures (Horowitz et al., 1979). The change across exposure exercises in presence and imagery vividness in the IE condition, was examined using multilevel modeling in $R$. Time (pre-treatment, post-treatment, or 1-month follow-up) was dummy-coded with pre-treatment as the reference group, to indicate each participant’s symptom change from pre-treatment at each time point. Models included fixed effects for each time contrast and each condition contrast, along with presence. Models also included all two-way interactions between time and condition, time and compliance, and treatment and compliance. Models also included all three-way interactions between time, condition, and compliance. Finally, models included fixed and random intercepts. Of particular interest were (1) the two-way interactions between condition and time, as well as (2) the three-way interactions between condition, time, and compliance. These parameters indicated (1) the association between condition and outcome and (2) whether compliance played a moderating role in the impact of condition. Significant interactions were probed with simple slope analysis. To facilitate interpretation, Cohen’s $d$ for each parameter was calculated as $\frac{2\left(\frac{Z}{\sqrt{N}}\right)}{\sqrt{1-{\left(\frac{Z}{\sqrt{N}}\right)}^2}}$, where $\text{Z}=\beta /SE$ (Rosenthal & DiMatteo, 2001). Cohen’s $d$ of 0.2, 0.5, and 0.8 denoted small, moderate, and large effects, respectively. All analyses were intent to treat. Missing outcome data (3.6%) were handled with full information maximum likelihood (FIML).

Results

Preliminary Analyses

Fisher’s exact test revealed no significant difference in gender across conditions, ${\chi }^2(2)=1.00$, p = .462, and ANOVA indicated no significant difference in age across conditions, $F(2,78)=1.00$, p = .657. Descriptive statistics for pretreatment and posttreatment symptoms are presented in Table 2. At pretreatment, there were no significant differences between the treatment and control groups on any symptom measure (all were $p>.05$). Descriptive statistics summarizing the anxiety composite score and self-efficacy data are presented in Table 3.

Table 2.

Summary of Pretreatment, Posttreatment, and Follow-up Measures by Condition

	SM			IE
	Pretreatment	Posttreatment	Follow-up	Pretreatment	Posttreatment	Follow-up
Measure	M (SD)	M (SD)	M (SD)	M (SD)	M (SD)	M (SD)
SIAS	50.61 (13.59)	47.50 (13.27)	47.33 (13.30)	51.77 (12.22)	31.04 (10.03)	30.82 (9.83)
SPDQ	19.03 (3.21)	18.41 (3.66)	18.74 (3.68)	19.03 (3.21)	11.55 (3.40)	11.49 (3.47)
GSES	27.30 (3.71)	27.63 (3.86)	27.93 (4.11)	27.22 (3.64)	31.33 (3.97)	31.36 (3.85)

Note: $N=82$. IE = imaginal exposure. SM = self-monitoring (control). M = mean. SD = standard deviation.

Table 3.

Summary of Means and Standard Deviations Among Anxiety Composite and Self-Efficacy Scores

		Pretreatment	Posttreatment	Follow-up
Anxiety	IE	0.98 (0.26)	0.32 (0.31)	0.34 (0.29)
	SM	0.95 (0.34)	0.97 (0.30)	0.95 (0.37)
Self-Efficacy	IE	0.54 (0.19)	0.79 (0.23)	0.76 (0.19)
	SM	0.58 (0.24)	0.59 (0.20)	0.58 (0.17)

Note: $N=82$. IE = imaginal exposure. SM = self-monitoring (control).

Descriptive Analyses

Calculated compliance rates were 59% for IE (requested thrice daily completion) and 62% for self-monitoring (requested 8 times daily completion), which were not significantly different from each other, $\beta =1.00$, $SE=1.00$, $Z=1.00$, $p=1.00$. There was no significant difference between self-monitoring ($M=1.00$; $SD=1.00$) and IE ($M=1.17$; $SD=0.72$), $\beta =0.54$, $SE=0.80$, $Z=0.68$, $p=0.12$ in reported number of social situations engaged in between prompts. However, the reported number of social situations avoided between prompts differed significantly by condition such that self-monitoring ($M=1.24$; $SD=0.56$) had more avoided situations on average than imaginal exposure ($M=0.92$; $SD=0.43$), $\beta =1.24$, $SE=0.44$, $Z=2.82$, $p=0.02$.

Additional descriptive analyses were conducted on data specific to the IE smartphone application. Automated engagement metrics were also tracked through the average total time (in minutes) spent in the smartphone application ($M=171.36$; $SD=35.72$) and the average number of logins ($M=18.39$; $SD=3.13$) across the duration of the study. Related to SUDS, the average first peak ($M=61.68$; $SD=12.81$) and average highest peak across any exposure exercise ($M=67.91$; $SD=15.33$) were collected. Finally, examining pooled fear hierarchy data revealed that on average 76.39% of hierarchy items were completed over the course of the study.

Social Anxiety

Please see Table 4 for model results. There was a significant two-way interaction between treatment condition and time from pre-treatment to post-treatment, $\beta =-0.83$, $SE=0.18$, $Z=-4.61$, $p=0.04$, $d=0.38$ and pre-treatment to 1-month follow-up, $\beta =-0.80$, $SE=0.22$, $Z=-3.63$, $p=0.03$, $d=0.37$ showing significantly greater change in social anxiety from IE compared to self-monitoring (see Figure 2). Additionally, there was a significant three-way interaction involving condition, compliance, and time from pre-treatment to post-treatment, $\beta =-0.98$, $SE=0.32$, $Z=-3.06$, $p=0.01$, $d=0.52$, and pre-treatment to follow-up, $\beta =-0.93$, $SE=0.35$, $Z=-2.66$, $p=0.01$, $d=0.52$, indicating that more compliance during the treatment predicted greater change in social anxiety.

Table 4.

Effect of Therapy Condition and Compliance on Outcome

	Social Anxiety Composite				Self-efficacy
	$\beta$	$SE$	$Z$	$d$	$\beta$	$SE$	$Z$	$d$
Intercept	0.96*	0.11	8.73	0.21	0.55*	0.09	6.11	0.19
Pre-Post	−0.74*	0.22	−3.36	0.24	0.63*	0.08	7.88	0.22
Pre-FU	−0.71*	0.20	−3.55	0.23	0.56*	0.09	6.22	0.21
Condition	−0.84**	0.24	−3.50	0.31	0.60*	0.32	1.88	0.22
Compliance	0.07	0.28	0.25	0.06	0.17	0.34	0.50	0.04
Condition*Pre-Post	−0.83*	0.18	−4.61	0.38	0.65*	0.19	3.42	0.33
Conditi on*Pre-FU	−0.80*	0.22	−3.63	0.37	0.64*	0.14	4.57	0.32
Compliance*Condition	−0.58**	0.30	−1.93	0.27	0.69*	0.26	2.65	0.22
Compliance*Pre-Post	0.19	0.26	0.73	0.05	0.12	0.12	1.00	0.04
Compliance*Pre-FU	0.26	0.22	1.18	0.05	0.14	0.18	0.78	0.04
Compliance*Condition*Pre-Post	−0.98**	0.32	−3.06	0.52	0.78*	0.25	3.12	0.35
Compliance*Condition*Pre-FU	−0.93**	0.35	−2.66	0.52	0.76*	0.26	2.92	0.35

Note. $N=82$. Condition = Self-monitoring vs IE. FU = follow-up.

^*p < .05

^**p < .01

^***p < .001.

Simple slope analysis indicated that in the IE condition, higher compliance was associated with significantly greater social anxiety symptom reduction from pre-treatment to post-treatment, $\beta =-0.92$, $SE=0.30$, $Z=-3.07$, $p=0.01$, $d=0.53$, and from pre-treatment to 1-month follow-up, $\beta =-0.91$., $SE=0.31$, $Z=-2.94$, $p=0.01$, $d=0.50$. Unlike in the IE condition, in the self-monitoring condition, higher compliance was not associated with symptom change in social anxiety from pre-treatment to post-treatment, $\beta =0.13$, $SE=0.19$, $Z=0.68$, $p=0.39$, $d=0.05$, or from pre-treatment to 1-month follow-up, $\beta =0.10$, $SE=0.19$, $Z=0.53$, $p=0.34$, $d=0.04$.

Self-Efficacy

There was a significant two-way interaction involving condition and time from pre-treatment to post-treatment, $\beta =0.65$, $SE=0.19$, $Z=3.42$, $p=0.04$, $d=0.33$, and pre-treatment to 1-month follow-up, $\beta =0.64$, $SE=0.14$, $Z=4.57$, $p=0.04$, $d=0.32$, indicating that IE led to greater increased self-efficacy than the self-monitoring control condition (see Figure 2). Furthermore, there was also a significant three-way interaction involving condition, compliance, and time from pre-treatment to post-treatment $\beta =0.78$, $SE=0.25$, $Z=3.12$, $p=0.04$, $d=0.35$, and pre-treatment to follow-up $\beta =0.76$, $SE=0.26$, $Z=2.92$, $p=0.04$, $d=0.35$, indicating a moderating effect of compliance on self-efficacy during the treatment.

Similarly, simple slope analysis indicated that in the IE condition, higher compliance was associated with significantly greater improvements in self-efficacy from pre-treatment to post-treatment, $\beta =0.74$, $SE=0.23$, $Z=3.22$, $p=0.04$, $d=0.23$, and from pre-treatment to 1-month follow-up, $\beta =0.75$, $SE=0.20$, $Z=3.75$, $p=0.04$, $d=0.23$. Unlike in the IE condition, in the self-monitoring condition, higher compliance was not associated with changes in self-efficacy from pre-treatment to post-treatment, $\beta =0.15$, $SE=0.19$, $Z=0.79$, $p=0.45$, $d=0.02$, or from pre-treatment to 1-month follow-up, $\beta =0.12$, $SE=0.24$, $Z=0.50$, $p=0.38$, $d=0.03$. Thus, in alignment with our expectations, compliance was associated with greater pre-treatment to post-treatment and follow-up social anxiety symptom change and self-efficacy in the IE condition but predicted not in the self-monitoring control.

Additionally, for variables only assessed in the IE condition, there was a significant increase in presence across time, $\beta =0.08$, $SE=0.05$, $Z=1.60$, $p=0.03$, $d=0.53$. Relatedly, imagery vividness also displayed a significant increase over time, $\beta =0.07$, $SE=0.04$, $Z=1.75$, $p=0.03$, $d=0.51$. Thus, both presence and imagery vividness improved with practice.

Discussion

This study examined the efficacy and effectiveness of 7-day IE exercises (vs. active control) on individuals with social anxiety related to social interactions. For this study, a neutral task involving self-monitoring of participants’ social avoidance behavior served as the active control condition. Descriptive results showed IE (59%) and self-monitoring (62%) had compliance rates that were not significantly different. However, those using the IE smartphone application displayed greater reductions in social anxiety than did self-monitoring. Similar benefits were found when examining self-efficacy. IE users showed significant increases at subsequent timepoints. Further, more completed IE exercises produced significantly greater reductions in social anxiety and greater increases in self-efficacy at subsequent timepoints compared with self-monitoring.

Well-designed technological tools can be an exciting avenue for advancing psychological interventions for anxiety, but they require sufficient engagement to realize their potential influence. Engagement among the IE intervention was strong, as the 59% compliance rate far exceeded the reported average (36%) in a meta-analytic review of smartphone intervention studies for anxiety (Linardon & Fuller-Tyszkiewicz, 2020). Paired with encouraging total time spent in the smartphone application and frequency of logins, this study suggests compelling evidence for the feasibility and acceptability of offering a smartphone application-based IE intervention for individuals with SAD. Significant utility of the smartphone application also allowed for adequate investigation of the impact of the intervention.

Many technological treatment tools have been developed without examining their efficacy (Alyami et al., 2017). This is the first study to our knowledge to use personal devices to deliver application-based stand-alone self-help IE for the treatment of SAD. Recent investigations exhibited important efficacy findings using smartphone applications for social anxiety, although they involved a mental health coach (Stolz et al., 2018) or were not specific to IE (Boettcher et al., 2018). Therefore, the current study contributes novel findings related to brief (one-week) IE-specific self-help exercises delivered for SAD using an accessible, technological medium.

Self-help cognitive and behavioral technological interventions have been shown to be cost-effective methods of delivering empirically validated treatments for a variety of psychological problems (Lewis et al., 2012; Newman et al., 2011). These tools can be valuable resources for those not currently in treatment. Such self-reliant tools may be especially important in exposure. From an emotional processing theory perspective (Foa & Kozak, 1986), it has been proposed that fear extinction does not entirely eliminate prior pathological associations among stimuli and responses but instead creates new associations to be retrieved more frequently than the pathological ones (Bouton, 2000). To achieve engagement with such tools, brevity may be a vital piece of intervention design (Fleming et al., 2018; Simblett et al., 2018; Torous et al., 2018). It may be that shortening interventions allows exposure interventions to streamline the creation, strength, and retrieval of non-threatening associations within one’s memory through repetition, and core tenets of inhibitory learning theory (Tolin, 2019). Shortening of interventions has been successful in prior treatment utilizing IE, as a 30-minute IE session was not significantly different from a 60-minute IE (van Minnen & Foa, 2006). Of note, the exercises delivered in this study were significantly shorter (i.e., 13 minutes) to be more adaptive for repeated completion throughout the day and were delivered without administration by a therapist. Additionally, prompting individuals to practice in vivo exposure following self-help IE could provide further compliance and improved outcome.

Smartphone self-help treatment tools may be valuable at various points in individuals’ treatment seeking status. Such tools may serve as an engaging, self-paced transition into psychological treatment. However, multicomponent platforms and applications may be overwhelming and ineffective for individuals at initial stages. For individuals who are currently receiving treatment from a mental health professional, IE may support and advance skills learned in therapy and could also serve as homework outside of therapy sessions. Such homework practice is likely to help anxious individuals to feel more confident, competent, and flexible at encountering feared stimuli across different contexts (Newman et al., 1997a). For those who have successfully completed therapy, smartphone application-based IE exercises could replace booster sessions which are shown to be beneficial (Gearing et al., 2013). Unfortunately, a substantial number of individuals experience a return of fear following exposure (Craske et al., 2014). Following successful exposure, relapses can occur where after an elapsed amount of time or change in context, the anxiety associated with the stimuli or situation returns following fear extinction (Shin & Newman, 2018; Vervliet et al., 2013). Therefore, focusing on the generalizability of social anxiety exposure by employing IE in various contexts could be beneficial in preventing anxious relapse through the repetitive reinforcement of non-pathological response structures (Foa & McNally, 1996; Huppert et al., 2003). Tools like the one in this study may assist with maintaining gains, while encouraging autonomy and self-efficacy.

Relatedly, the relationship between social anxiety and self-efficacy could be better understood. The observed changes in self-efficacy were certainly encouraging. In the context of imagined skill building, such changes were expected based on prior literature noting the relationship between change in self-efficacy predicting change in anxiety (Bandura & Adams, 1977; Bandura et al., 1980). In future studies, such findings could be replicated by collecting self-efficacy measures through EMA.

The current study was the first one to test a technology-delivered IE exercise. Findings illuminate the rich opportunity that exists from a variety of technological mediums. Notably, presence was observed to increase over time, supporting a novel finding made in a recent study involving VRE (Zainal et al., 2021). Relatedly, imagery vividness also displayed significant increases over time. Additional investigation into the differences in presence during IE and VRE could be valuable, given that one involves imaginal, mental imagery and the latter involves virtual or video images. However, our finding preliminarily suggests that repeated exposure practice can increase the sense of involvement, realness, and spatial presence in IE. It has been noted that resulting symptom and behavior changes are dependent largely on the visceral sensory experience of the individual (Lang, 1977). Participants were encouraged to conduct in vivo exposure homework in addition to the use of their IE exercises. Since traditional exposure treatment encourages in vivo homework practice but most VRE studies do not, investigating the contribution of in vivo exposure in the success of the IE condition would be informative. These important findings may suggest a necessary revisiting of the constructs of presence and mental imagery ability as less static than previously understood. Further, there are implications for anxiety research and treatment. However, with repeated practice of imagery-related tasks, there is initial evidence to suggest individuals may have the ability to develop and improve their imagery skills.

There are several limitations to the current study. First, participants were limited to a diagnosis of SAD. Future studies utilizing exposure and IE techniques in advanced and personalized ways should investigate if our findings are replicated and extended to other disorders. In particular, it may be beneficial to target disorders that often utilize exposure-based practices, such as specific phobias or PTSD. Similarly, the current study examined comparisons between a smartphone-based exposure technique and an active, self-monitoring control. Thus, future studies should examine whether our findings are replicated and extended to comparisons of smartphone-based exposure interventions against more traditional delivery of exposure in the context of face to face therapy. From a technological standpoint, a direct comparison to VRE would also be useful. Another limitation was the short-term nature of the 1-month follow-up. Subsequent studies could test the limits of the current intervention’s lasting effects. Furthermore, our understanding of the findings could have been improved with addition of measures of app acceptability, as well as credibility and expectancy of the treatment. Though the findings are encouraging, diversity in the current study was limited within our sample and future studies should examine whether such techniques improve outcome in more diverse groups. Relatedly, the use of undergraduate students to comprise the sample may not be representativeness of general population. Therefore, a larger sample size could lead to more expansive demographic characteristics to better understand the generalizability of the findings. Finally, since the study was conducted during the COVID-19 pandemic, there were restrictions to opportunities to extend IE practice to consistent in vivo practice in social situations.

The IE smartphone application in the current study was isolated from face-to face therapy or other co-occurring treatment. Future studies could offer this self-help treatment tool or other advanced interventions to see if they augment therapies by increasing interest, compliance to assignments, promote skill building, and further improve outcome. It could be that blended therapy (i.e., a limited number of psychotherapy sessions with a therapist in addition to self-guided material) may be the optimal format and modality to integrate advancing technological treatment tools (e.g., Newman et al., 1997b; Newman et al., 2014). Because the study trial length was relatively brief, especially in comparison to typical CBT involving exposure, there may be a benefit in extending the access and use of the exercises.

When comparing IE exercises delivered via a novel smartphone application, we found that those individuals had greater reductions in social anxiety and avoidance, as well as increases in self-efficacy. Findings suggest that the self-help IE exercises delivered via a smartphone application are effective in targeting important aspects of social anxiety. There are clinical implications that may require revisiting the expected length of SAD exposure interventions. The current study was a brief yet, fairly intensive delivery of the treatment. Brief repeated interventions in routine clinical care may be able to produce similar benefits as would be found with standard length treatments. Importantly, developing an understanding of individual differences that contribute to or hinder improvements during exposure and IE could alter the course of exposure therapy.

Highlights.

• Implemented a novel smartphone application treatment for social anxiety disorder

• Participants using the application experienced greater reductions in social anxiety

• Those using the application also experienced greater increases in self-efficacy

• Higher treatment compliance was associated with greater changes in outcome