Introduction
Traditional in-person psychotherapies are unable to address mental health needs around the world (Kohn et al., Reference Kohn, Saxena, Levav and Saraceno2004). Among the mental health conditions, social anxiety disorder (SAD), also known as social phobia, is one of the most prevalent mental disorders in Western countries (Bandelow and Michaelis, Reference Bandelow and Michaelis2015). It has an early onset, follows a chronic course if not treated, and is associated with a significant impairment in quality of life (Carpenter et al., Reference Carpenter, Curtiss, Hofmann, McKay, Abramowitz and Storch2017). Although the prevalence of SAD in China (0.2–0.7% current prevalence, and 4.11% lifetime prevalence; Guo et al., Reference Guo, Meng, Huang, Fan, Zhou, Ling and Su2016; Shen et al., Reference Shen, Zhang, Huang, He, Liu, Cheng and Kessler2006) is lower than in many Western countries (0.6–8% 12-month prevalence and 2.8–13% lifetime prevalence; Bandelow and Michaelis, Reference Bandelow and Michaelis2015), it still translates into an enormous number of people based on the large population in China.
Internet-based interventions are a potentially feasible solution to closing the large gap between mental health treatment need and actual treatment delivery, especially in low- and middle-income countries (Arjadi et al., Reference Arjadi, Nauta, Chowdhary and Bockting2015). In the last decade, internet-based cognitive behaviour therapy (ICBT) for SAD, which is a highly promising approach to increasing accessibility and availability to treatment, has been developed and validated in Western countries (Andersson, Reference Andersson2009; Berger et al., Reference Berger, Caspar, Richardson, Kneubuhler, Sutter and Andersson2011; Titov et al., Reference Titov, Andrews, Schwencke, Robinson, Peters and Spence2010) as well as in China (Kishimoto et al., Reference Kishimoto, Krieger, Berger, Qian, Chen and Yang2016). SAD is one of the disorders for which ICBT has the strongest empirical support (Boettcher et al., Reference Boettcher, Carlbring, Renneberg and Berger2013; Hedman et al., Reference Hedman, Botella, Berger, Lindefors and Andersson2016). However, a substantial proportion of patients do not achieve clinically significant improvement. In studies on ICBT for SAD, clinically significant change rates range between 40% and 60%, which means approximately half of the participants did not achieve significant improvements (Boettcher et al., Reference Boettcher, Carlbring, Renneberg and Berger2013). The few studies on predictors of treatment outcomes in ICBT for SAD yielded inconsistent findings (Hedman et al., Reference Hedman, Botella, Berger, Lindefors and Andersson2016). As such, it remains unclear which patients and patient characteristics are associated with benefiting most from this particular treatment delivery method.
Some of the factors that have been found to be associated with treatment outcomes in ICBT include demographic (e.g. age, gender, educational level and employment status) and clinical (e.g. severity of symptoms and comorbid diagnosis) characteristics (El Alaoui et al., Reference El Alaoui, Ljotsson, Hedman, Kaldo, Andersson, Ruck and Lindefors2015; Hedman et al., Reference Hedman, Andersson, Ljotsson, Andersson, Andersson, Schalling and Ruck2012; Nordgreen et al., Reference Nordgreen, Havik, Ost, Furmark, Carlbring and Andersson2012). However, findings are inconsistent. For example, employment status (i.e. working full-time) was associated with positive outcomes in ICBT for SAD in one study (Hedman et al., Reference Hedman, Andersson, Ljotsson, Andersson, Andersson, Schalling and Ruck2012), but not in another (El Alaoui et al., Reference El Alaoui, Ljotsson, Hedman, Kaldo, Andersson, Ruck and Lindefors2015). In terms of clinical characteristics, one study found that higher intensity of baseline SAD symptoms predicts more symptom improvement (Nordgreen et al., Reference Nordgreen, Havik, Ost, Furmark, Carlbring and Andersson2012) while another study found the opposite (Hedman et al., Reference Hedman, Botella, Berger, Lindefors and Andersson2016). Similarly, although comorbid depression was associated with lower treatment response in one study of ICBT for SAD (Hedman et al., Reference Hedman, Andersson, Ljotsson, Andersson, Andersson, Schalling and Ruck2012), another study reported no association between comorbid depression and treatment outcomes (Nordgreen et al., Reference Nordgreen, Havik, Ost, Furmark, Carlbring and Andersson2012). These equivocal results may be due to differences in populations, treatments and outcome measures (e.g. diagnosis-free status, clinical significant improvement, post-treatment level of social anxiety). However, differences across studies may also be due to differences in treatment engagement, including attrition (drop-out) and adherence.
Treatment adherence is one of the most important issues faced by developers of web-based CBT interventions. Importantly, adherence to ICBT predicted treatment outcomes in several studies (El Alaoui et al., Reference El Alaoui, Ljotsson, Hedman, Kaldo, Andersson, Ruck and Lindefors2015; Hedman et al., Reference Hedman, Botella, Berger, Lindefors and Andersson2016; Hilvert-Bruce et al., Reference Hilvert-Bruce, Rossouw, Wong, Sunderland and Andrews2012). However, some studies do not report a clear association between treatment adherence and outcomes (Nordgreen et al., Reference Nordgreen, Havik, Ost, Furmark, Carlbring and Andersson2012; Schulz et al., Reference Schulz, Stolz, Vincent, Krieger, Andersson and Berger2016). In internet-based treatments, adherence refers to the extent to which individuals make use of the content of the internet intervention (Christensen et al., Reference Christensen, Griffiths and Farrer2009), and is often operationalized as the number of modules or exercises completed (Beatty and Binnion, Reference Beatty and Binnion2016). An inter-related but conceptually distinct construct is attrition or drop-out, which is often operationalized as failing to complete the research trial protocol and assessments (Christensen et al., Reference Christensen, Griffiths and Farrer2009). Many users of internet-based interventions do not complete the intervention in its entirety (Beatty and Binnion, Reference Beatty and Binnion2016; Farrer et al., Reference Farrer, Griffiths, Christensen, Mackinnon and Batterham2014). Consequently, adherence in ICBT needs to be examined not only as an outcome predictor but also as an outcome itself, whose predictors need to be identified and investigated.
Research has shown certain demographic and clinical factors to be associated with adherence in ICBT (Batterham et al., Reference Batterham, Neil, Bennett, Griffiths and Christensen2008; El Alaoui et al., Reference El Alaoui, Ljotsson, Hedman, Kaldo, Andersson, Ruck and Lindefors2015; Newby et al., Reference Newby, Mewton, Williams and Andrews2014; Nordgreen et al., Reference Nordgreen, Havik, Ost, Furmark, Carlbring and Andersson2012; van Straten et al., Reference van Straten, Cuijpers and Smits2008; Williams et al., Reference Williams, O’Moore, Mason and Andrews2014). However, again, studies have yielded equivocal results possibly due to differences in sample sizes, participant characteristics, definitions of adherence, treatment modality and targeted symptoms. For example, some studies found that females (Batterham et al., Reference Batterham, Neil, Bennett, Griffiths and Christensen2008; El Alaoui et al., Reference El Alaoui, Ljotsson, Hedman, Kaldo, Andersson, Ruck and Lindefors2015) and older individuals (Newby et al., Reference Newby, Mewton, Williams and Andrews2014; Williams et al., Reference Williams, O’Moore, Mason and Andrews2014) were more likely to adhere to ICBT, whereas others have found that gender and age were unrelated to adherence (Newby et al., Reference Newby, Mewton, Williams and Andrews2014; van Straten et al., Reference van Straten, Cuijpers and Smits2008; Williams et al., Reference Williams, O’Moore, Mason and Andrews2014), or even that younger individuals were more adherent (Batterham et al., Reference Batterham, Neil, Bennett, Griffiths and Christensen2008; van Straten et al., Reference van Straten, Cuijpers and Smits2008). Occupational status was consistently unrelated to adherence in ICBT for SAD (El Alaoui et al., Reference El Alaoui, Ljotsson, Hedman, Kaldo, Andersson, Ruck and Lindefors2015; van Straten et al., Reference van Straten, Cuijpers and Smits2008) as well as other symptoms (Farrer et al., Reference Farrer, Griffiths, Christensen, Mackinnon and Batterham2014; Spek et al., Reference Spek, Nyklicek, Cuijpers and Pop2008). However, there have been mixed results on the relationship between level of education and adherence (Batterham et al., Reference Batterham, Neil, Bennett, Griffiths and Christensen2008; El Alaoui et al., Reference El Alaoui, Ljotsson, Hedman, Kaldo, Andersson, Ruck and Lindefors2015; van Straten et al., Reference van Straten, Cuijpers and Smits2008). In terms of clinical symptoms, pre-treatment social anxiety symptom severity does not typically predict adherence in ICBT (El Alaoui et al., Reference El Alaoui, Ljotsson, Hedman, Kaldo, Andersson, Ruck and Lindefors2015; Newby et al., Reference Newby, Mewton, Williams and Andrews2014; Nordgreen et al., Reference Nordgreen, Havik, Ost, Furmark, Carlbring and Andersson2012; van Straten et al., Reference van Straten, Cuijpers and Smits2008; Williams et al., Reference Williams, O’Moore, Mason and Andrews2014). Nonetheless, one study of a large community-based sample reported greater adherence among individuals with higher baseline anxiety symptom severity (Batterham et al., Reference Batterham, Neil, Bennett, Griffiths and Christensen2008).
A number of studies on ICBT have tested the efficacy of different approaches to enhance adherence (Berger et al., Reference Berger, Caspar, Richardson, Kneubuhler, Sutter and Andersson2011; Hilvert-Bruce et al., Reference Hilvert-Bruce, Rossouw, Wong, Sunderland and Andrews2012). A common strategy used is therapist support or guidance. Several meta-analyses show superiority of therapist-guided versus self-guided ICBT in terms of drop-out rates, adherence to treatment and efficacy (Baumeister et al., Reference Baumeister, Reichler, Munzinger and Lin2014; Richards and Richardson, Reference Richards and Richardson2012; Spek et al., Reference Spek, Cuijpers, Nyklicek, Riper, Keyzer and Pop2007). However, some studies have reported no significant differences between self- versus therapist-guided ICBT on treatment adherence and outcomes in social anxiety (Berger et al., Reference Berger, Caspar, Richardson, Kneubuhler, Sutter and Andersson2011; Dear et al., Reference Dear, Staples, Terides, Fogliati, Sheehan, Johnston and Titov2016). One important discussion related to this strategy focuses on identifying the minimum effective dose of support and guidance needed to enhance not only treatment adherence, but also treatment outcomes (Newman et al., Reference Newman, Erickson, Przeworski and Dzus2003). More research is needed to understand how treatment adherence is related to treatment outcomes in ICBT, as well as the variables associated with improved adherence. The present study contributes to this body of literature. Specifically, the aim of the present study was to explore the effects of treatment type (unguided versus low-intensity therapist support), clinical population (participants with social anxiety disorder versus subclinical SAD features) and demographics (age, gender, education, occupation, income) on the outcomes of drop-out (failure to complete post-treatment assessments), adherence (average number of modules completed) and social anxiety (Social Interaction Anxiety Scale and Social Phobia Scale residual gain scores) in an 8-week self-guided online CBT intervention for social anxiety among a sample of adults (ages 18–45 years) recruited online from the general population in China.
Method
The current study is a continuation of a study which examined the efficacy of ICBT (versus a wait-list control group) for social anxiety (Kishimoto et al., Reference Kishimoto, Krieger, Berger, Qian, Chen and Yang2016). In the study, participants were assigned to a therapist-guided ICBT condition, a self-guided ICBT condition, or to a wait-list control condition. Self-report social anxiety outcome measurements were completed at pre- and post-treatment. Of relevance to the current study, a significant interaction effect of condition by time was found, with both active treatments (therapist-guided and self-guided ICBT) being superior to a waiting-list control condition. With regard to improvements in social anxiety symptoms and the mean number of completed modules, no significant differences were found between the therapist-guided and self-guided treatment conditions.
Treatments included in the current analysis were conducted between October 2013 and October 2015. All participants were recruited online from the general population. Prior to treatment, participants gave informed consent. The study was approved by the Ethics Committee for Protecting Human and Animal Subjects of Peking University (no. 2013-10-18). This trial was retrospectively registered at the Chinese Clinical Trial Registry (ChiCTR1900021670).
Participants
A total of 255 participants were recruited via a two-step screening procedure as depicted in Fig. 1. First, advertisements for the opportunity to participate in the study, which provided general information about SAD and ICBT, were posted on several local university listservs, professional psychology websites and social media platforms (e.g. WeChat and WeiBo). Interested potential participants could fill in an online questionnaire including demographic information, the Chinese versions of the Social Phobia Scale (SPS) and Social Interaction Anxiety Scale (SIAS; Mattick and Clarke, Reference Mattick and Clarke1998; Ye et al., Reference Ye, Qian, Liu and Chen2007) to sign up for the study. Individuals who met the following inclusion criteria were then invited to the second step of the screening process: (1) at least 18 years old; (2) exceeding cut-off values on the SPS (cut-off ≥ 26) or SIAS (cut-off ≥ 32); (3) not taking psychiatric medications in the past year; (4) not receiving psychological treatment in the past 6 months; and (5) computer literacy and reliable access to the internet during the study.Footnote 1 The second step of screening involved completing the Chinese version of the Mini International Neuropsychiatric Interview (MINI; Si et al., Reference Si, Shu, Dang, Su, Chen, Dong and Zhang2009) administrated via a telephone call. Inclusion criteria for this step of screening were: (6) reporting no history of suicide risk; (7) either (a) having no any diagnosis of mental disorder according to MINI or (b) having a diagnosis of SAD; and (8) if diagnosed with SAD, having no comorbid diagnoses besides anxiety disorders, major depression and/or dysthymia on the MINI. Finally, participants were assigned to either self-guided (N = 183) or therapist-guided (N = 72) ICBT.
Screening and allocation of participants.
Treatment and procedure
Although this study was first planned as a randomized controlled trial, we chose to abandon the randomization procedure due to the magnitude of the mismatch between public demand for the treatment and therapist capacity. All participants received access to the same ICBT intervention programme, which was translated and culturally adapted into Chinese (Kishimoto et al., Reference Kishimoto, Krieger, Berger, Qian, Chen and Yang2016) from a well-established ICBT protocol (Berger et al., Reference Berger, Caspar, Richardson, Kneubuhler, Sutter and Andersson2011). It consists of eight text-based lessons completed over 8 weeks, each with a specific topic and practical exercises. Therapists were clinical psychology graduate students at Peking University supervised by T.K. and M.L. Participants in the therapist-guided ICBT condition could email their therapist if they had questions about the content of the intervention or if they had difficulties understanding or applying the skills. Also, if participants did not make weekly progress in the programme, their therapist would email to encourage them to continue.
Dependent variables
Study drop-outs
Drop-out was defined as failure to complete outcome assessments based on the intent-to-treat sample size of the intervention group. Participants who were allocated to the intervention but never started the intervention were also counted as drop-outs.
Adherence to treatment
Adherence was operationalized as number of completed modules of the intervention, ranging from 0 to 8. This information was recorded online at the administration website of the ICBT programme. A module was considered as completed if the participant (a) ticked all the items in the checklist at the end of the module, indicating completion of reading materials and relevant exercises, and (b) added at least one valid entry in the homework section of the module.
Outcome
The Chinese versions of the SPS and SIAS were used as the primary outcome measures (Mattick and Clarke, Reference Mattick and Clarke1998; Ye et al., Reference Ye, Qian, Liu and Chen2007). These measures assess social anxiety-related concerns about being scrutinized or judged during routine activities, as well as fear in social interaction situations. The SPS contains 20 items, is rated on a 0–4 scale, and yields total scores between 0 and 80. The SIAS contains 19 items, is rated on a 0–4 scale, and yields total scores between 0 and 76. Both scales have good psychometric properties, including high internal consistency (Cronbach’s α was 0.90 for SPS and 0.87 for SIAS in the current sample) and good test–retest reliability (0.85 for SPS and 0.86 for SIAS; Ye et al., Reference Ye, Qian, Liu and Chen2007).
Both the SPS and SIAS were assessed through online questionnaires at both pre- and post-treatment. For the analyses, residual gain scores were calculated, which account for measurement error of repeated administration of the instruments and the initial differences between individuals at pretreatment (Steketee and Chambless, Reference Steketee and Chambless1992). The residual gain scores were calculated using the formula z 2 − (z 1 × r 1,2), where z 2 is the Z-transformed post-treatment score and z 1 is the transformed pre-treatment score, and r 1,2 is the Pearson correlation between pre- and post-assessments. Residual gain scores were reversed so that higher scores would indicate greater improvement.
Potential predictors
Clinical characteristics
Social anxiety disorder diagnosis, comorbid major depression or dysthymia, and comorbid anxiety disorder were candidate clinical predictors of drop-out and adherence in the current study. Comorbid major depression or dysthymia and comorbid anxiety disorders were included as candidate predictors of treatment outcomes, while social anxiety disorder diagnosis was excluded, as pretreatment social anxiety levels were controlled for utilizing residual gain scores.
Demographic characteristics
Demographic data were collected during the screening of participants. The following demographic characteristics were investigated as potential predictors of drop-out, adherence and outcome: age, gender, occupational status, educational level and monthly income.
Statistical analysis
Statistical analyses were conducted using SPSS version 20.0. Linear and logistic regression analyses were performed adopting the two-step approach proposed by de Graaf et al. (Reference de Graaf, Hollon and Huibers2010), by first identifying significant single univariate predictors, and subsequently adding those simultaneously into a final multiple regression model. In the first step, the models were built in two blocks, where the first block was the predictor and the second block was treatment group condition (self-guided versus therapist-guided) and the interaction term (group × predictor). Potential moderators were investigated if the interaction term was significant. In the second step, significant predictors from the initial univariate analyses were analysed together in a final multiple regression model using backward deletion.
Results
Participants
Participant characteristics for both groups are presented in Table 1. Only occupational status differed significantly between the two groups.
Description of participants and group differences
ICBT, internet-based cognitive behaviour therapy; RMB, Renminbi, Chinese Yuan; SIAS, Social Interaction Anxiety Scale; SPS, Social Phobia Scale.
Predictors of drop-out and adherence
No significant group difference was found for the odds of drop-out (OR 0.802, p = .43), with 60% (109/183) of participants in the self-guided ICBT and 54% (39/72) of participants in the therapist-guided ICBT dropping out. However, those diagnosed with social anxiety disorder were significantly less likely to drop out (OR 0.531, p = .03).
Participants completed 4.1 modules (SD 3.3) on average. No group differences were found for the number of completed modules (t 253 = 1.04, p = .30). Average completed modules were 4.0 (SD 3.2) for the self-guided ICBT and 4.4 (SD 3.5) for the therapist-guided ICBT, respectively. Figure 2 depicts the non-usage attrition curve. Plotted are number of completed modules against the proportion of remaining participants completing them.
Percentage of module completion.
The first-step analysis showed that there were no significant moderators (p > .05). Univariate linear regression analyses yielded age and SAD diagnosis as significant predictors of adherence. None of the other investigated predictors (gender, occupational status, educational level, monthly income) reached significance, although occupational status approached significance (see Table 2).
Univariate regression analyses predicting number of completed modules
In the second-step analysis, the multiple regression model retained age (B = 0.17, SE = 0.04, p = .008) and SAD diagnosis (B = 0.16, SE = 0.44, p = .01) as significant predictors of adherence (R 2 = .05, F = 7.13, p = .001). Older participants and participants with a diagnosis of SAD tended to complete more modules.
Predictors of residual gain scores of SIAS and SPS
In the first-step analysis, no significant moderator was found (p > .05). Univariate linear regression analyses yielded gender and number of completed modules as significant predictors of residual gain score for SIAS. Age and a comorbid anxiety disorder diagnosis approached significance for SPS scores (see Table 3).
Univariate regression analyses predicting residual gain score of SIAS/SPS
SIAS, Social Interaction Anxiety Scale; SPS, Social Phobia Scale.
In the second-step analysis, the model retained gender (B = −0.20, SE = 0.18, p = .04) and number of completed modules (B = 0.24, SE = 0.03, p = .01) as significant predictors of residual gain score for SIAS (R 2 = .11, F = 6.35, p = .002). Participants who identified as female and those who competed more modules reported greater improvement in SIAS scores.
Discussion
Predictors of treatment adherence and outcomes were investigated in the context of self-guided and therapist-guided ICBT for social anxiety in China. We examined whether therapist support resulted in improved treatment adherence or outcomes, namely to test the moderation effect of intervention condition. No significant effects were found for therapist guidance predicting treatment drop-out, the number of completed modules or SIAS/SPS scores. These results are consistent with some (Berger et al., Reference Berger, Caspar, Richardson, Kneubuhler, Sutter and Andersson2011; Dear et al., Reference Dear, Staples, Terides, Fogliati, Sheehan, Johnston and Titov2016; Furmark et al., Reference Furmark, Carlbring, Hedman, Sonnenstein, Clevberger, Bohman and Andersson2009), but not all previous studies (Beatty and Binnion, Reference Beatty and Binnion2016; Richards and Richardson, Reference Richards and Richardson2012).
One possible explanation is that the intensity of therapist guidance provided in our study was insufficient to enhance outcomes. For example, most emails were not personalized, therapists were students, and the average amount of time spent responding to participants was approximately 15 minutes per week per participant. However, the evidence to support these explanations are also mixed. Firstly, 15 minutes of therapist contact per week is not uncommon in ICBT studies for anxiety disorders (see review: Apolinario-Hagen, Reference Apolinario-Hagen2019). Moreover, previous reviews reported no impact of the therapist qualification (Baumeister et al., Reference Baumeister, Reichler, Munzinger and Lin2014). However, our sample was also relatively young (about 25 years old on average). In other countries, this age group has found to be more open to self-guided online interventions (March et al., Reference March, Day, Ritchie, Rowe, Gough, Hall and Ireland2018). Their increase in mental health related self-help seeking behaviours online, are partly due to concerns of stigma and autonomy (Rickwood et al., Reference Rickwood, Deane and Wilson2007). Given that individuals of this age group may tend to self-select to engage in self-guided ICBT for a sense of reduced stigma and increased autonomy, it would not be surprising if our sample had lower intentions of using therapist-guided ICBT, thus leading to a non-significant difference between the two conditions. In sum, more research is necessary to determine what type of support and the level of support are needed to achieve significant and clinically meaningful differences in treatment engagement and outcomes.
With regard to predictors of outcomes, neither comorbid depression/dysthymia nor other anxiety disorders were predictive of treatment outcomes, which replicates results from previous studies of predictors of ICBT for SAD (Nordgreen et al., Reference Nordgreen, Havik, Ost, Furmark, Carlbring and Andersson2012; Schulz et al., Reference Schulz, Stolz, Vincent, Krieger, Andersson and Berger2016). Both women and participants with greater adherence (completing more modules) demonstrated the greatest improvements in SIAS scores, but not SPS scores. Age (older participants) and comorbid anxiety disorders approached significance as predictors of improvement in SPS scores (p < .06), but not SIAS scores. Although the two scales are usually highly correlated, they do in fact measure different aspects of social anxiety. The SPS assesses fears of being scrutinized during routine activities (eating, drinking, writing, etc.), while the SIAS assesses fears of more general social interaction (Mattick and Clarke, Reference Mattick and Clarke1998). The difference in outcome measures might be a possible explanation of equivocal results also found in previous studies (El Alaoui et al., Reference El Alaoui, Ljotsson, Hedman, Kaldo, Andersson, Ruck and Lindefors2015; Hedman et al., Reference Hedman, Botella, Berger, Lindefors and Andersson2016; Nordgreen et al., Reference Nordgreen, Havik, Ost, Furmark, Carlbring and Andersson2012).
In the present study, SAD diagnosis was associated with lower drop-out and higher adherence rates. Although several studies found pre-treatment social anxiety symptom severity to not be predictive of adherence to ICBT (El Alaoui et al., Reference El Alaoui, Ljotsson, Hedman, Kaldo, Andersson, Ruck and Lindefors2015; Newby et al., Reference Newby, Mewton, Williams and Andrews2014; Nordgreen et al., Reference Nordgreen, Havik, Ost, Furmark, Carlbring and Andersson2012; van Straten et al., Reference van Straten, Cuijpers and Smits2008; Williams et al., Reference Williams, O’Moore, Mason and Andrews2014), our results were in line with one earlier study composed of a larger community-based sample (Batterham et al., Reference Batterham, Neil, Bennett, Griffiths and Christensen2008). One possible explanation is that those with less social anxiety symptoms may have experienced less benefit from the intervention (due to floor effects) and thus dropped out. On the other hand, participants with SAD diagnosis might be more motivated to improve than those without SAD diagnosis, and thus were more adherent. Previous research has shown there to be a significant correlation between motivation and adherence (Al-Asadi et al., Reference Al-Asadi, Klein and Meyer2014).
Older age was associated with better adherence. Evidence from previous studies was mixed, with almost an equal number of studies finding either older age or younger age to be associated with higher adherence (Batterham et al., Reference Batterham, Neil, Bennett, Griffiths and Christensen2008; Newby et al., Reference Newby, Mewton, Williams and Andrews2014; van Straten et al., Reference van Straten, Cuijpers and Smits2008; Williams et al., Reference Williams, O’Moore, Mason and Andrews2014). In a recent review of predictors of adherence (Beatty and Binnion, Reference Beatty and Binnion2016), the authors identified discrepancies between what were considered ‘older’ versus ‘younger’ participants across different studies to be driving the inconsistent findings related to the association between age and adherence. According to their findings, adults over 25 years old had the highest adherence rates. These findings were consistent with our own, as the mean age of our sample was 25.6 years.
In previous research on ICBT for SAD (El Alaoui et al., Reference El Alaoui, Ljotsson, Hedman, Kaldo, Andersson, Ruck and Lindefors2015; van Straten et al., Reference van Straten, Cuijpers and Smits2008) and other symptoms (Farrer et al., Reference Farrer, Griffiths, Christensen, Mackinnon and Batterham2014; Spek et al., Reference Spek, Nyklicek, Cuijpers and Pop2008), occupational status was not related to adherence. However, in our study, student status approached significance. That is, non-student participants showed a tendency to be more adherent than students. This finding makes sense, considering the fact that students in China have more access to mental health services than non-students (Qian et al., Reference Qian, Chen, Zhang and Zhang2010). Accordingly, non-student participants might have valued the intervention more, as they may have perceived mental health resources to be scarcer. Alternatively, given that older participants were more adherent in general, it could also be that personality traits and occupational factors in relation to individual traits (e.g. more self-regulation and higher conscientiousness with growing age) might be moderating variables. The interaction between those variables awaits further examination.
Several limitations of this study should be acknowledged, which may also help outline opportunities and directions for future research. Firstly, drop-out (failure to complete outcome assessment) rate was high in our study. Although it is not unusual for ICBT studies to have high drop-out rates (Beatty and Binnion, Reference Beatty and Binnion2016), it limits the reliability and validity of our results regarding the predictors of drop-out, adherence and outcomes. More research is needed in order to understand the best strategies for recruiting and retaining subjects in online interventions in non-Western, low- and middle-income contexts. Secondly, it should be noted that our findings are based on a relatively small and homogenous sample. Future studies with larger samples and more variability in terms of demographics, clinical characteristic and therapy processes may be able to identify other variables associated with treatment adherence and outcomes. Thirdly, the therapist-guided ICBT (N = 183) and self-guided ICBT (N = 72) conditions were unequally sized due to a failure to maintain randomization. Thus, the interpretation of our results are limited and caution should be taken when comparing our findings with others. Fourthly, our study examined a small number of the many possible predictors of drop-out, adherence and outcome. It was beyond the scope of our study to examine all the predictors that have been identified in the literature. Thus, investigating additional factors such as treatment credibility, chronicity of symptoms and earlier experiences of treatment are warranted. Lastly, follow-up data were not included in our study and thus we were not able to examine predictors of long-term effectiveness of ICBT for social anxiety. As an earlier study indicated that predictors for short-term and long-term improvements might be different (de Graaf et al., Reference de Graaf, Huibers, Riper, Gerhards and Arntz2009), future studies should include more follow-up data and identify factors associated with maintenance of symptom improvement.
Finally, this study was only exploratory in nature, serving to generate hypotheses for clinicians and researchers. Researchers have pointed to the potential of a range of other theoretical models to increase adherence focusing on technology (e.g. persuasive system design; Kelders et al., Reference Kelders, Kok, Ossebaard and Van Gemert-Pijnen2012) or the individual, including the health belief model (Rosenstock et al., Reference Rosenstock, Strecher and Becker1988), the protection motivation theory (Maddux and Rogers, Reference Maddux and Rogers1983), the theory of reasoned action (Ajzen and Fishbein, Reference Ajzen and Fishbein1980), the theory of planned behaviour (Ajzen, Reference Ajzen, Kuhl and Beckman1985), the social-cognitive theory (Bandura, Reference Bandura1986), and models based on self-efficacy (Bandura, Reference Bandura1997). However, a coherent theory-driven, evidence-based model for understanding treatment engagement in online interventions is needed. For example, the behaviour change model for internet interventions describes an integrated process for how effective internet interventions produce and maintain behaviour changes and symptom reductions (Ritterband et al., Reference Ritterband, Thorndike, Cox, Kovatchev and Gonder-Frederick2009). That is, user characteristics (e.g. disease, demographics, traits, beliefs and attitudes) affect their website use and adherence, which, in turn, are influenced by website characteristics and the support offered by the programme. Website use further leads to behaviour changes and symptom improvement. Incorporating such a model into both research and clinical practice could potentially help researchers and clinicians evaluate and develop individualized and targeted interventions. Meanwhile, some researchers have adopted qualitative methods to obtain detailed and individual feedback from participants in order to gain a more thorough understanding of participants’ experience. For example, insufficient support due to the absence of a therapist and the lack of specificity of the contents to one’s own problems were identified as two common reasons for drop-out of a transdiagnostic online intervention (Fernandez-Alvarez et al., Reference Fernandez-Alvarez, Diaz-Garcia, Gonzalez-Robles, Banos, Garcia-Palacios and Botella2017). Studies exploring motivations to persist with ICBT suggest offering supporting conditions such as enhancing a sense of control and autonomy and building identification with and trust in the programme (Donkin and Glozier, Reference Donkin and Glozier2012; Wilhelmsen et al., Reference Wilhelmsen, Lillevoll, Risor, Hoifodt, Johansen, Waterloo and Kolstrup2013). More qualitative research is needed in this area, as such qualitative studies help generate hypotheses for what can predict and enhance treatment outcome and adherence.
In conclusion, being female and higher adherence (i.e. more modules completed) was associated with better treatment outcomes. Older participants and those with a diagnosis of SAD reported higher adherence. However, the low-intensity therapist support provided in this study did not yield significant improvements in treatment adherence or outcomes. Research is needed to explore support using different providers (e.g. peers), mediums (e.g. video chat), formats (e.g. group), intensity (e.g. higher frequency and duration of contact) or foci (e.g. validation or motivational interviewing).
Acknowledgements
The authors would like to acknowledge Professor Dr Alexander L. Gerlach and Dr Yang Sun for their kind guidance and support in the writing of the manuscript.
Financial support
This work was supported by National Social Science Foundation of China (project number 15ZDB139); China Scholarship Council (H.C.); and the James B. Duke International Travel Fellowship (M.R.).
Conflicts of interest
The authors declare that they have no conflicts of interest.
Ethical statements
The authors have abided by the Ethical Principles of Psychologists and Code of Conduct as set out by the APA (http://www.apa.org/ethics/code/). The study was approved by the Ethics Committee for Protecting Human and Animal Subjects of Peking University (no. 2013-10-18).
Comments
No Comments have been published for this article.