記住我
Stress is rapidly becoming a major issue affecting adults in high-income countries, especially during periods of uncertainty and worry. Chronic stress is closely related to mental illnesses such as anxiety disorders and depression, leading to various symptoms such as sleep disturbances, pain, dizziness, cardiovascular and digestive problems, as well as fatigue [,]. Younger individuals, particularly students [-], are experiencing a decline in mental health on a global scale [,]. Studies indicate that approximately 11% of students experience impairments such as anxiety, depression, exhaustion, and burnout-like symptoms [,]. Furthermore, a high level of stress can have a negative impact on academic performance, resulting in changes in study direction, prolonged studies, and even dropout [,].
Students encounter distinct challenges during their academic journey, including the need to assimilate a substantial amount of content, effectively manage their time, cope with performance expectations, and handle examination pressure []. In addition, the developmentally sensitive period associated with this age group, combined with the academic environment, can contribute to increased stress levels []. Furthermore, compared to previous generations, today’s students appear to exhibit lower stress tolerance and inadequate stress coping mechanisms, which further exacerbate the situation [,,]. Notably, a recent study by Ehrentreich et al [] reported that stress levels among students have increased by nearly 40% due to the impact of the COVID-19 pandemic.
To prevent students from experiencing chronic stress and its long-term effects, the implementation of appropriate prevention programs is crucial. These programs aim to promote students’ self-management and stress management skills, including learning and time management techniques, to help them effectively cope with stress and to counteract increasing stress levels in the target group [,]. Studies have demonstrated the positive impact of interventions such as behavioral therapy–based approaches, relaxation and mindfulness exercises, psychoeducation, and time and study management strategies in reducing stress among students [,,]. Typically, evidence-based stress management programs combine psychoeducational sessions with relaxation exercises [-]. Importantly, stress management programs should be specifically tailored to the needs of students. By considering the target group’s real-life context, these programs facilitate the transfer of acquired skills into everyday life [].
Despite the importance of stress management programs for students, successful uptake remains challenging []. Unfortunately, individuals experiencing stress often do not make use of stress management techniques for several reasons. These include the fear of being stigmatized [], underestimation of the impact of stress, limited availability of therapy options, and high cost, particularly for young people in education [,].
Low-threshold, mobile health (mHealth) interventions such as smartphone apps could potentially bridge this gap. A meta-analysis by Weisel et al [] highlighted the advantages of apps, including location and time independence, reduced stigmatization, and low costs []. Initial evidence suggests that smartphone apps can effectively reduce perceived stress, distress, depression, and anxiety and improve quality of life, psychological health, well-being, and self-regulation among student populations [-]. However, reported disadvantages of digital interventions, such as low adherence, legal concerns, lack of therapist relationship, and arbitrary scheduling, may diminish their effectiveness [,].
Conversational agents (CAs), commonly known as chatbots, are designed to simulate humanlike conversations and are increasingly used in clinical and nonclinical settings [-]. Initial findings demonstrate the feasibility, acceptance, and effectiveness of CAs in various health domains [,], including promoting physical activity []; managing pain []; reducing substance abuse [,]; improving depression, distress, and stress []; enhancing general wellness and pain []; and facilitating self-adherence and psychoeducation []. Although large language model (LLM)–based CAs have recently gained increasing attention [], they are still subject to basic research in computer science because of several severe shortcomings, such as hallucinations and nonconscious bias, among others []. Therefore, LLM-based CAs are not yet appropriate for safe and ethical delivery of several-week health interventions []. Hence, we decided to implement an established, safe, and transparent approach to using CAs and used a rule-based CA [,,-].
Studies investigating the effectiveness of stress management interventions delivered by a CA specifically tailored to the needs of students are still lacking. While recent studies have explored interventions such as Stressbot, developed with Meta’s Messenger (Meta Platforms, Inc) and CA Atena, accessible via Telegram messaging app (developed by the Digital Health Lab at Fondazione Bruno Kessler FBK research center), their focus has been limited to short-term outcomes or specific topics. For instance, while Stressbot aimed to reinforce coping self-efficacy, its intervention period was only 7 days []. Similarly, CA Aetna’s positive psychology and cognitive behavioral approaches with a tailored focus on the unique needs of the COVID-19 pandemic rather than the life context of students led to inconclusive outcomes regarding anxiety and stress reduction []. Furthermore, a previous study evaluating an artificial intelligence (AI)–based chatbot that provided self-help interventions for students to reduce depression lacked detailed descriptions of evidence-based intervention designs, leaving uncertainty about the elements implemented []. However, evidence-based design is vital in developing CA-based coaching intervention programs [] and stress management interventions for specific groups such as students []. To our knowledge, there is no study describing the development and evaluation of the effectiveness of a CA-delivered stress management coaching program lasting several weeks and adapted to the specific context of students in their everyday lives.
Consequently, we have developed an evidence-based, scalable, and CA-delivered stress management coaching intervention for students called MISHA. It combines the following components: (1) providing psychoeducation about stress, mindfulness, and relaxation; (2) fostering participant motivation for self-reflection on stress and stress reactions; and (3) guiding participants in the regular practice of mindfulness and relaxation techniques. This comprehensive approach addresses key aspects of stress management, including knowledge acquisition, self-reflection, and practical application of mindfulness and relaxation techniques [,]. By focusing on these evidence-based intervention components, MISHA aims to empower students with effective tools and strategies to reduce stress and its long-term effects.
ObjectivesThe goal of this pilot study was twofold: (1) to develop a scalable, evidence-based coaching intervention specifically designed for students and delivered via a CA and (2) to assess the coaching intervention’s effectiveness, engagement, and acceptance.
MISHA was developed in collaboration with the ETH Zurich using the open-source software platform MobileCoach [], designed for rule-based digital health interventions [,-]. MISHA features a chat-based interface with multimedia elements and regular notifications to engage users. The app includes a chat channel, an audio library with relaxation exercises, psychoeducational illustrations, and frequently asked questions (). Communications takes place via predefined but dynamic answer options or by providing free-text input. Study participants were provided with access to a beta version of the MISHA app for Android (Google LLC) devices through Firebase [] and for iOS (Apple Inc) devices through TestFlight [].
Figure 1. Screenshots of the MISHA app (coach selection, chat interface, reminder, and audio library). Translation from German to English, screenshot Select coach: "Choose a coach"; screenshot Chat with coach: "Effective time management can support you and prevent or reduce stress. Shall we discuss this?", "Yes, I’m interested.", "Great, you’re on board. Today, we’ll focus on reflecting on your personal thought and behavior patterns related to time management. Remember, time management is primarily self-management.", "Really?", "Perhaps you’ve experienced this yourself or observed it in others…"; screenshot Reminder: "Have you relaxed today? See you tomorrow", "Dear Isabelle, tomorrow I’ll show you a relaxation exercise”; screenshot Audio library: “Progressive Relaxation - Introduction (long)", "Progressive Relaxation - Brief", "Progressive Relaxation - Extended", "Seated Meditation", "Footprints in the Snow", "Waterfall”. Coaching Concept of MISHAThe intervention concept for MISHA draws inspiration from an effective face-to-face prevention program [], adapting its content and topics to suit a CA-delivered approach. MISHA’s chat messages and notifications are aligned with the health action process approach (HAPA) model, emphasizing both motivational and volitional processes in behavior change [].
MISHA integrates evidence-based strategies from cognitive behavioral therapy (CBT), mindfulness, and psychoeducation to provide information about stressors and coping techniques [,]. The stress management program includes fundamental elements derived from CBT, such as cognitive restructuring, identification, evaluation, and modification of maladaptive thought patterns []. In addition, techniques such as behavioral activation and activity monitoring from CBT were applied to directly support the participants in their desired goals in a collaborative approach. For further details on CBTs and session elements, refer to . The overall aim is to empower participants to reflect on their daily stressors and effectively manage their stress with new coping techniques.
Coaching ContentMISHA offers a consecutive 12-session coaching program based on the stress management manual by Kaluza []. Sessions cover psychoeducation on stress, relaxation techniques, and student-specific topics such as examination anxiety. Topics are personalized, for example, setting goals, individual appointments with the CA, or selecting a CA. Participants can schedule sessions every 2 to 4 days, completing the program in 24 to 54 days (refer to for an overview of sessions and a detailed description of the content). Throughout the coaching, participants receive personalized feedback on the progression of the coaching, motivational reminders, and reminders in case of inactivity (refer to for detailed information on reminders). Personalization on an individual level is essential in promoting trust, engagement, adherence, and effectiveness to digital health interventions [,].
Study Design and ProcedureWe conducted an unblinded, 2-armed, pilot randomized controlled trial in a population of university students in Switzerland. Study participants were allocated either to a 4-week to 7-week coaching intervention or to a 40-day waitlist control group. This research project was registered at the German Clinical Trials Register accredited by the World Health Organization (DRKS00030004). The trial was conducted following CONSORT-EHEALTH (Consolidated Standards of Reporting Trials of Electronic and Mobile Health Applications and Online Telehealth) guidelines. No significant content changes were made to the coaching intervention during the study period.
After downloading the MISHA app, participants were greeted and provided with information about the study procedure and coaching program. They were explicitly informed that the app does not serve as a substitute for psychotherapy and were given guidance on where to seek further help if needed. Study information was displayed within the app. To proceed, participants had to provide electronic informed consent by confirming that they had read and understood the study information. Subsequently, inclusion criteria were checked, and participants were directed to the baseline self-assessment at preintervention (time point 1; T1) using the app’s in-built LimeSurvey platform (LimeSurvey Project). The MobileCoach software automatically randomized participants into either the intervention or the waitlist control group by a 1:1 allocation using random numbers (0 to 1), with numbers <0.5 assigned to the intervention group. Participants from the intervention group started the coaching program immediately. Upon program completion (1) by working through all the modules or (2) after 54 intervention days, participants were directed to the postintervention survey (time point 2; T2) before moving to the final goodbye session. During the intervention, further self-reported outcomes (eg, goal achievement) and use data (eg, total minutes spent on in-app relaxation) were gathered.
Participants from the waitlist control group received short weekly chat messages from MISHA, informing them about the remaining duration of their wait and encouraging them to continue their participation in the study. After 40 days of waiting, they were presented with the postintervention survey (T2) and given the opportunity to participate in the coaching program.
There was no human involvement throughout the study; however, participants had the option to contact the study team via email if they encountered technical issues or encountered problems with app download.
Ethical ConsiderationsThe Cantonal Ethics Committee of Zurich (KEK-ZH, BASEC-Nr. Req-2020-01038) reviewed the research project and confirmed that the study did not fall within the scope of the Human Research Act. All participants gave informed electronic consent by selecting a checkbox before enrolling in the study and were informed about their right to opt out at any time. Their data were deidentified. Participants who completed the postintervention survey had the opportunity to win a voucher worth CHF 200 (US $224.73). In addition, students of applied psychology at Zurich University of Applied Sciences had the opportunity to earn 5 test person hours.
RecruitmentFrom October 6, 2021, to the end of October 2021, flyers were distributed via email to students at the University of Zurich, the Zurich University of Teacher Education, University of Applied Sciences Northwestern Switzerland School of Education, the University of Teacher Education in Special Needs Zurich, and the Zurich University of Applied Sciences. In addition, the flyer was posted on Facebook (Meta Platforms, Inc) and LinkedIn (Microsoft Corp). The app could be downloaded via flyer by following a web link. Eligibility was determined within the MISHA app by self-report and included the following: (1) being aged ≥18 years; (2) possession of and basic knowledge in the use of a smartphone; (3) sufficient knowledge of the German language; and (4) being a student at a Swiss university, university of applied sciences, university of teacher education, or college of higher education.
OutcomesPrimary OutcomeTo measure the effectiveness of the program, we assessed perceived stress at preintervention (T1) and postintervention (T2) time points using the German version of the Perceived Stress Scale, a self-report questionnaire consisting of 10 items []. Participants rated their responses on a scale ranging from 0 (never) to 5 (very often).
Secondary OutcomesWe measured secondary outcomes, including depression, anxiety, somatic symptoms, and active coping, at preintervention and postintervention time points by self-report. presents all outcomes and time points.
Depression, Anxiety, and Somatic SymptomsWe used the Patient Health Questionnaire Somatic, Anxiety, and Depressive Symptom Scales [] to detect depressive symptoms, anxiety, and somatic symptoms, which consists of the Patient Health Questionnaire-9 (PHQ-9), Generalized Anxiety Disorder-7, and the Patient Health Questionnaire-15. The PHQ-9 is a 9-item questionnaire assessing depressive symptoms []. Participants rate the frequency of each symptom over the past 2 weeks, ranging from 0 (not at all) to 3 (nearly every day). The Generalized Anxiety Disorder-7 is a 7-item questionnaire that measures anxiety symptoms []. Participants rate the frequency of each symptom over the past 2 weeks, ranging from 0 (not at all) to 3 (nearly every day). The Patient Health Questionnaire-15 is a 15-item questionnaire measuring psychosomatic symptoms []. Participants rate the severity of each symptom over the previous 4 weeks, ranging from 0 (not bothered at all) to 2 (bothered a lot). For this study, items 14 (trouble with sleeping) and 15 (ie, low energy or tiredness) were collected in the PHQ-9 (similar in both questionnaires) but had to be converted according to the manual []. By combining these individual components, the PHQ Somatic, Anxiety, and Depressive Symptoms Scales provide a comprehensive assessment of depressive symptoms, anxiety, and somatic symptoms.
Active CopingAccording to the HAPA model [], we evaluated participants’ engagement in stress management activities by asking them to rate how often they had actively taken steps to reduce stress in the past 5 days. The question was assessed on a rating scale ranging from 1 (never) to 4 (regularly). This allowed us to understand the participants’ level of proactive involvement in managing their stress.
Predictor: Self-Efficacy ExpectancyVarious health behavior change models, including the HAPA model [], consider self-efficacy expectancy to be a key aspect of health behavior change. However, research findings on the impact on stress interventions are mixed [-]. To address this, we assessed self-efficacy expectancy using the General Self-Efficacy Scale []. Before the intervention, participants rated their agreement with statements on their ability to handle tasks effectively on a 4-point Likert scale ranging from 1 (not at all true) to 4 (exactly true). The total score of the General Self-Efficacy Scale ranges from 10 to 40, with higher scores indicating higher self-efficacy.
ExploratoryWorking AllianceTo assess the interaction between participants and MISHA, we used the German version of the Working Alliance Inventory-Short Revised [] after the intervention. This self-report questionnaire comprises 12 items that capture the quality of the therapeutic relationship and collaboration between participants and the CA via 3 dimensions: goal, task, and bond. Responses were rated with an adapted scale from 1 (I do not agree at all) to 6 (I completely agree) after the intervention.
Subjective Stress Expertise and Goal AchievementThroughout the coaching period, we assessed participants’ goal achievement 3 times (sessions 1, 6, and 11) using a scale of 1 to 10, where 1 referred to the goal as clearly not achieved and 10 referred to the goal as fully achieved. We further measured participants’ stress expertise 3 times (sessions 2, 5, and 13) using a similar scale, ranging from 1 (no idea how stress manifests itself in me) to 10 (I know exactly how I react when under stress).
Engagement and AcceptanceThe extent to which a participant has to engage with the intervention to derive the maximum benefits is termed intended use []. For MISHA, we defined intended use for participants as completing the postintervention assessment, regardless of completing all sessions. This definition was based on the fact that participants may have varied goals and desired outcomes, leading to differences in their use of MISHA’s features, including frequency and duration [,]. It also implies that participants do not necessarily need to interact with all available intervention components. Furthermore, participants might discontinue using the intervention upon achieving their personal goals, indicating that nonuse is not due to loss of interest [,]. In addition, we ground this approach on the self-determination theory, where autonomy by providing choice is essential [].
To assess participants’ engagement in the coaching program, we analyzed use data from the intervention group by calculating the ratio of replied conversational turns based on the number of SMS text messages sent by MISHA in relation to SMS text messages replied by participants. Furthermore, we tracked the number of sessions completed by participants and the number of reminders sent to participants in cases of inactivity (ie, if participants stopped interacting during a session). In addition, we tracked the number of minutes of audio files played by participants throughout the intervention.
We evaluated the feasibility and acceptance of MISHA using the user version of the Mobile App Rating Scale (uMARS) [] after the intervention. The uMARS is a validated questionnaire that assesses the dimensions of engagement, functionality, esthetics, information, perceived quality, and perceived impact. All subscales use a 5-point Likert scale ranging from 1 to 5, where higher scores indicate a more favorable judgment. In this study, 19 items were translated from English to German to assess engagement (eg, entertainment, interest, customization, interactivity, and target group of the app), information (eg, quality of information, quantity of information, visual information, and credibility of source), perceived quality (eg, recommendation, use, payment, and overall rating), and perceived impact (eg, awareness, knowledge, attitudes, behavior change, seeking help, and intention to change). In addition to the uMARS, participants had the opportunity to provide feedback in free text prompted by the following questions: “What did you like most about the MISHA app?” and “What would you improve in the MISHA app?”
Sample Size CalculationThe sample size was estimated for a generalized estimating equation (GEE) based on a repeated-measure (within-between interaction) ANOVA. A small to medium time by group interaction effect size (Cohen f=0.15) for the primary outcome perceived stress due to prior results [] was expected. The G*Power (Heinrich-Heine-Universität Düsseldorf) analysis [] revealed that a sample size of 90 participants would be sufficient with a power of 0.80 and a correlation of r=0.5 between measurements. Owing to the high percentage of dropouts observed in earlier studies, the target sample was increased to 180 participants [].
Data AnalysisDescriptive statistics, independent 2-tailed t tests, and chi-square tests were conducted to analyze baseline differences in demographics and outcomes between the intervention and control groups.
In our analysis, we examined the effectiveness of the intervention by assessing changes in the primary outcome perceived stress scores over time within each group (intervention and control) and comparing these changes between groups. We first conducted a per-protocol (PP) analysis, including only participants who completed both surveys. This was done using a repeated-measure ANOVA with perceived stress as the dependent variable, time as the within-subject factor, and group as the between-group factor. Secondary outcomes, including depression, anxiety, psychosomatic symptoms, and active coping, were analyzed accordingly.
In compliance with the CONSORT (Consolidated Standards of Reporting Trials) guidelines, we also conducted an intention-to-treat (ITT) analysis wherein all randomized participants were included, regardless of their adherence to the coaching intervention. This analysis was performed using GEE. In model 1, we conducted an unadjusted evaluation with time (T1 and T2), group (intervention and control), and treatment (group by time interaction) as independent variables, with perceived stress as the dependent variable. The incorporation of time allows the examination of the dependent variable stress over different time points, the incorporation of group allows for comparison of stress between groups, and the interaction between group and time allows for an examination of whether the changes in outcomes over time differ between the intervention and control groups. In model 2, we did an adjusted analysis with the inclusion of the covariate general self-efficacy for the primary outcome perceived stress. The same independent variables were considered as in model 1. Secondary outcomes were evaluated accordingly. A log link function, gamma distribution, and unstructured covariance structure were applied. This modeling approach provided the best fit with the outcomes and allowed us to avoid restrictions on the covariance structure. To reduce the impact of influential observations and outlier effects, we used a robust estimator, which is consistent with standard procedures when using GEE.
Using GEE [] offered several advantages. First, it allowed us to consider the correlations between the measurement times in longitudinal data, which is important for analyzing repeated measures. In addition, GEE allowed us to include incomplete data sets using an estimating equation to handle missing data. GEEs use all available data and estimate missing outcome values under the assumption of missing completely at random (MCAR). To assess the assumption of MCAR, we conducted the Little MCAR test. Calculations of between-group effect sizes (Cohen d) were based on the pooled SD and labeled as small (Cohen d=0.2), medium (Cohen d=0.5), and large (Cohen d=0.8). Furthermore, we explored the potential relation of working alliance and perceived impact on treatment outcomes using a correlation. All statistical analyses were performed using SPSS software (version 28; IBM Corp). We applied qualitative content analysis [,] using thematic maps [] to answer the open-ended questions.
In total, 230 individuals downloaded the app. Of the 230 individuals, 148 (64.3%) were assessed for eligibility and completed the baseline survey. Before randomization, of the 148 participants, 8 (3.5%) discontinued using the app and 140 (60.9%) were randomized into intervention (70/140, 50%) and waitlist control (70/140, 50%) groups. The complete participant flow is depicted in .
Participants had a mean age of 26.71 (SD 6.29) years. While 23.6% (33/140) of the participants identified as men, 73.6% (103/140) as women, and 2.1% (3/140) as nonbinary, 0.7% (1/140) declined to provide information about their gender (). Regarding relationship status, 59.3% (83/140) of the participants reported being married or in a relationship, while 40.7% (57/140) were single. Regarding educational background, most participants (90/140, 64.3%) had an apprenticeship or vocational or high-school diploma. A substantial proportion of the participants (37/140, 26.4%) had a university degree at the bachelor level or higher vocational education or training, while 8.6% (12/140) had other qualifications. Regarding their field of study, most participants (131/140, 93.6%) were studying at a university of applied sciences or university, while 5% (7/140) were studying at other institutions. The participants had a degree in (applied) psychology (124/140, 88.6%), social sciences (6/140, 4.4%), or other fields (7/140, 5%). There were no differences between groups for any of the outcomes at baseline.
Figure 2. Study flowchart. ITT: intention-to-treat; PP: per-protocol; T1: time point 1. Table 1. Sample description at baseline (n=140).OutcomeControl group (n=70)Intervention group (n=70)P valueaAge (y), mean (SD)26.21 (5.56)27.22 (6.96).75Gender, n (%).78aBaseline group comparison between intervention group and waitlist control group with t test or chi-square test. Italicized values are statistically significant.
bBSc: Bachelor of Science.
cPSS-10: Perceived Stress Scale-10.
dPHQ-9: Patient Health Questionnaire-9.
eGAD-7: Generalized Anxiety Disorder-7.
fPHQ-15: Patient Health Questionnaire-15.
gGSES: General Self-Efficacy Scale.
hHAPA: health action process approach.
EffectivenessTo evaluate the effectiveness of the intervention and to take missing data into account, a PP analysis of the time by group interaction was conducted followed by an ITT analysis. For the PP analysis (), we found evidence of a treatment effect (group by time interaction) from pre- to postintervention time points between the intervention and control groups for stress (P=.001; Cohen d=−0.60), depressive symptoms (P=.003; Cohen d=−0.50), and psychosomatic symptoms (P=.010; Cohen d=−0.36) but not for anxiety and active coping behavior.
In the ITT analysis for the unadjusted model (model 1), we found evidence of a treatment effect (group by time interaction) from pre- to postintervention time points between the intervention and control groups for stress (P<.001), depressive symptoms (P=.003), and psychosomatic symptoms (P=.003). No treatment effect was found for anxiety (P=.13) and active coping (P=.09).
After adjusting for the covariate self-efficacy expectancy (model 2), we found evidence of treatment effect sizes similar to model 1 (). Furthermore, there was evidence for an effect of self-efficacy expectancy on perceived stress (P<.001), depression (P<.001), anxiety (P<.001), and psychosomatic symptoms (P<.001) but not on active coping.
Table 2. Preintervention and postintervention means, results of the per-protocol (PP) repeated-measure ANOVA analysis, and between-group effect sizes (Cohen d) of primary and secondary outcomes (n=98).MeasurePreintervention, mean (SD)Postintervention, mean (SD)Between-group effect sizes (intervention group vs waitlist control group after the intervention)aCohen d values based on means and the pooled SD of the PP analysis.
b95% CI of Cohen d (between groups, after the intervention).
cPSS-10: Perceived Stress Scale-10.
dNot applicable.
ePHQ-9: Patient Health Questionnaire-9.
fGAD-7: Generalized Anxiety Disorder-7.
gPHQ-15: Patient Health Questionnaire-15.
hHAPA: health action process approach.
Table 3. Results of the outcome intention-to-treat analysis (model 1), including self-efficacy as covariate (model 2), using generalized estimating equations.OutcomeModel 1aModel 2baModel 1: unadjusted model (without covariate).
bModel 2: adjusted model for general self-efficacy expectancy.
cPSS-10: Perceived Stress Scale-10.
dNot applicable.
eTime effect represents the rate of improvement for both the intervention and waitlist control groups.
fGroup effect represents intervention or waitlist control group.
gTreatment effect is represented by group and time interaction.
hPHQ-9: Patient Health Questionnaire-9.
iGAD-7: Generalized Anxiety Disorder-7.
jPHQ-15: Patient Health Questionnaire-15.
kHAPA: health action process approach.
ExploratoryRegarding the working alliance, participants in the intervention group reported a mean working alliance score of 4.23 (SD 0.89) after the intervention. When exploring the potential influence of the working alliance on changes in outcomes from pre- to postintervention time points, we did not find evidence for correlations on any of the outcomes (Pearson correlation r ranging from −0.021 to 0.223). The participants rated their subjective stress expertise and goal achievement throughout the coaching program (3 times). For goal achievement, we observed a significant increase from the first to the third measurement with a large effect size (Cohen d=−1.07). provides further details.
Table 4. Means for subscales bond, task, and goal of working alliance and results of a paired t test for stress expertise and goal achievement.aWithin group comparison: start of intervention versus end of intervention.
bWAI-SR: Working Alliance Inventory with Likert scale ranging from 1 to 7.
cNot applicable.
Engagement and AcceptanceIn the intervention group, 60% (42/70) of the participants finished the coaching program by completing the postintervention survey (completers) and used the intervention as intended. Although the Little test indicated that values were MCAR for perceived stress (χ21=0.5; P=.47), depression (χ21=0.2; P=.63), anxiety (χ21=2.0; P=.16), psychosomatic symptoms (χ21=0.6; P=.80), and active coping (χ21=0.1; P=.82), we conducted a dropout analysis due to the potential risk of differential attrition, particularly with significantly higher dropouts observed in the intervention group []. The analysis revealed no significant differences in outcomes (eg, stress and depression) or demographics (ie, gender and age) between completers and dropouts.
Overall, 45% (19/42) of the completers worked through all 13 sessions, played a mean of 86.52 (SD 120.54) minutes of relaxation audios, and received a mean of 115.88 (SD 5.06) reminders; provides further information. On average, MISHA sent 400 (SD 205.61) SMS text messages and participants answered a mean of 297.54 (SD 169.80) SMS text messages, resulting in an average engagement ratio of 74.3%.
The participants in the intervention group (42/70, 60%) rated the subscale information highest, with a mean of 4.26 (SD 0.46), followed by engagement (mean 3.42, SD 0.70), perceived impact (mean 3.35, SD 0.87), and subjective quality of the app (mean 2.99, SD 0.87). Regarding engagement, individual customization was rated lowest with a mean of 2.71 (SD 0.84), while the target group fit was perceived as high (mean 3.95, SD 0.90). The participants liked the visual information of the CA and rated it high regarding correctness, clarity, and logic (mean 4.45, SD 0.55). Only a few participants (2/42, 2%) showed a high willingness to pay for the app (mean 2.10, SD 0.91) or anticipated high future use (mean 2.98, SD 1.05). The recommendation of the app to others was good, with a mean of 3.43 (SD 1.19) within the subjective app quality scale.
Table 5. Indicators of engagement: intend
留言 (0)