Impact of an App-Based Resilience Training on Enhancing Stress Regulation and Mental Health

Human life is filled with challenges, from every-day stressors to significant adversities. Navigating these obstacles successfully is where resilience is required. Resilience is the remarkable ability to adapt quickly to changing environmental circumstances, maintaining or regaining mental health and functionality during or after adverse events [1‒5]. Central to resilience is the capacity for effective stress regulation [1, 6] enabling individuals to manage stress responses and emerge stronger from life's inevitable challenges. Understanding and enhancing resilience is essential for promoting mental well-being and thriving amidst adversity.

Due to its influence on well-being, health, and quality of life, a large body of research has focused on measuring resilience (for a review, see [7, 8]). While there are different subjective (self-assessment questionnaires) and objective (external assessments) approaches to capture an individual’s resilience [9], no “gold standard” has been found yet [8].

Resilience is known to increase with age [10]. This highlights the need to consider age as a significant factor in resilience studies. Additionally, sex-specific differences in stress regulation and resilience have been reported, with women generally exhibiting stronger skills in these areas [4], as well as the level of education [11]. Therefore, our study will pay particular attention to these differences when analyzing the effects of the intervention.

In recent years, digital tools have emerged as promising avenues for enhancing resilience. These tools offer non-invasive, continuous, easily accessible, and autonomous training opportunities. Utilizing artificial intelligence and machine learning algorithms, they can individualize resilience training to meet users’ specific needs (e.g., [12, 13]). Such innovations provide real-world context for improving stress-coping mechanisms and resilience. In addition to that, there are numerous studies, which found that resilience training has a beneficial effect on stress and mental health (e.g., [14, 15]). This effect seems to persist in different samples (e.g., children [16], acute [17] and chronic diseases [18, 19], college students [20]) and settings (workshops [21], web-based interventions [22, 23], mobile-based apps [16]).

In recent years, there has been an exponential growth in mobile technologies aimed at improving various mental health issues, as reported in a recent meta-analysis by Lecomte et al. [24]. These technologies, known as mobile health, encompass health support, including mental health, through mobile devices [25], and are considered a subset of digital health [26]. Linardon et al. [27] provided a comprehensive review of smartphone-based interventions for mental health problems, including a meta-analysis of randomized controlled trials. The results convincingly showed that smartphone interventions significantly outperformed control conditions in improving depressive symptoms, generalized anxiety symptoms, stress levels, quality of life, general psychiatric distress, social anxiety symptoms, and positive affect. Most effects were robust, even after adjusting for several possible confounding factors (type of control condition, risk of bias assessment). In sum, smartphone interventions were not significantly different from active interventions (face-to-face treatment, computer-assisted treatment); however, they offer the advantages of being easily accessible (since almost everyone owns a smartphone) and straightforward to use. Although mental health apps are not intended as a substitute for professional clinical services, the present findings highlight the potential of apps as a cost-effective, easily accessible, and low-intensity intervention for people who cannot receive standard mental health treatment [27].

The field of mobile health services is booming with between 10,000 and 20,000 mental health apps currently available [28], but it is estimated that only about 3–4% is evidence-based. In general, only a small proportion of these technologies have been subjected to empirical evaluation [29, 30].

One notable mental health app that has been supported by specific evidence-based studies is MoodMission [31]. Users were shown to experience a decrease in depression and improved self-efficacy in coping strategies. eQuoo [32], a gamified mobile intervention to improve resilience, teaches users psychological concepts such as emotional imperatives, generalization, and reciprocity through psychoeducation and storytelling, and achieves 90% adherence through gamification. iFightDepression [33] is an Internet-based example of a depression self-management tool that offers both self-administered tests and a guided online self-management program to help people with mild to moderate depression self-manage their symptoms.

Current papers on resilience training demonstrate positive results regarding stress reduction. A recent study using Enhanced Stress Resilience Training (ESRT) for UK Surgical Trainees has shown that participation in ESRT is associated with reduced stress and improved mindfulness. Additionally, ESRT has been beneficial for well-being and executive function. Additionally, the training also led to reduced burnout rates [34]. Another study found that a digital resilience intervention, called RESIST, can be effective training for promoting resilience and stress reduction, demonstrating moderate to large effects [35]. A study focusing on resilience training for the management of critical situations has also indicated that the training resulted in improved stress control as well as enhanced physical, behavioral, and/or cognitive performance [36].

The primary aim of this study was to investigate the effectiveness of an app-based resilience training program. We hypothesized that participants using the AI-Refit app will exhibit higher resilience scores post-training compared to those who do not use the app (hypothesis 1). Additionally, we expect that app users will improve their stress-coping strategies and achieve better outcomes than the control group (hypothesis 2). We further hypothesized that these enhanced resources will positively impact mental health by reducing mood disturbances, psychosomatic complaints, and anxiety symptoms (hypothesis 3). Finally, we hypothesize that older adults and women will benefit more significantly from the app-based training than younger adults and men (hypothesis 4). By exploring these hypotheses, this study aims to contribute valuable insights into the development of digital resilience training tools, ultimately enhancing mental health and well-being across diverse populations.

This study was conducted using an online setting (LimeSurvey version 6.4.1), which was designed by the Clinical Division of Psychiatry and Psychotherapeutic Medicine, Medical University of Graz (MUG). Participants were recruited via a recruiting company (probando.io) and a pool of healthy study participants at the Medical University of Graz. All participants gave written informed consent before they participated in the study, which was conducted under the Declaration of Helsinki and was approved by the Ethics Committee of the Medical University of Graz (EC-number: EK 49-297 ex 22/23).

The following self-report questionnaires were presented within the online survey in German language: Demographic Questionnaire – information on age, sex, and education was collected; Brief Symptom Inventory – 18 (BSI-18; [37]): to measure psychological distress and symptoms throughout the previous 7 days, the German version of the BSI-18 [38] was administered. It encompasses 18 items, which are evenly assigned to three subscales (depression, anxiety, somatization; range 0–24 for each scale). Items are rated on a five-point Likert scale, ranging from 0 (“not at all”) to 4 (“always”). The global severity index (GSI), which indicates global psychological distress, was created by calculating the sum of all items (score range = 0–72). Resilience Scale – 13 (RS-13, [39]): to measure resilience, the RS-13 was administered. It encompasses 13 items, which are rated on a seven-point Likert scale and range from 1 (“strongly disagree”) to 7 (“strongly agree”). A total score is constructed by summing up all item scores, with higher scores indicating higher levels of resilience. Stress and Coping Inventory (SCI; [40]): the SCI assesses current stress load and symptoms as well as how stress is dealt with using five coping strategies. In total, 54 items are assigned to five scales measuring current stress load (stress due to uncertainty, stress due to overload, stress due to loss and actual negative events, total stress, physical and psychological stress symptoms) and five stress-coping strategies (positive thinking, active stress coping, social support, keeping faith, increased alcohol, and cigarette consumption). Each scale consists of four items, which are answered on a four-point Likert scale, ranging from 1 (“do not agree at all”) to 4 (“strongly agree”). Item scores are summed up to create scale scores, with higher values representing higher stress load or better stress coping (except for the scale “increased alcohol and cigarette consumption”), in which higher values represent more maladaptive coping. For this study, the corresponding scales were summed up to create a total stress score and a total coping score. System Usability Scale (SUS; [41]): the SUS is used to evaluate the usability of technological systems. In the current study, it was presented after the resilience training (i.e., at t2 and t3), to measure the usability of the app. The SUS consists of ten items, which are rated on a five-point Likert scale, ranging from 1 (“strongly disagree”) to 5 (“strongly agree”). A composite measure of the overall system usability can be calculated by summing up the score contributions of each item and multiplying this sum by 2.5 (score range = 0–100).

As part of the development of a resilience-training app named AI-Refit, the MUG created a state-of-the-art training manual. This manual is based on the “7 pillars of resilience” [42]. Each pillar is divided into 5 sub-points consisting of psychoeducation, a reflection exercise, a mindfulness exercise, a relaxation exercise and a feedback section. In the psychoeducation section, the pillar itself is described and scientifically proven. In the reflection exercise, statements are made about the respective pillar itself, which may or may not apply to the reader and are labeled accordingly. In the two relaxation and mindfulness exercises, the client is offered various techniques or exercises that can strengthen or better support the respective pillar (e.g., self-awareness) for themselves. In the feedback section, the opinion on the “effectiveness” of each pillar can be documented and self-assessed.

Together with psychologists from the MUG and software experts from the JOANNEUM Research (JR) Graz, a jointly agreed concept was then developed for the implementation of an app based on this manual (see Fig. 1). The app was implemented by JR in Android Unity for use on tablets. The interaction design and the design of the graphical user interface were coordinated by the MUG. The participants were trained in the use of the app and had the opportunity to ask questions about content or technical problems at any time. They were asked to use the app as often as possible over the period of 2 months. There was no specification as to which training component should be executed as well as completed at which timepoint. At the end of each training component, they were asked whether they had completed the training on their own. This ensures that the content has been sufficiently completed by the person.

This study was designed as a randomized controlled trial (RCT). Participants were randomly assigned to either intervention group 1 (IG1) or intervention group 2 (IG2) using a random number generator in Excel. Participants were contacted by the study administration and provided with detailed information and instructions about the study. Those interested were informed about the study by a psychologist and gave written consent to participate. Participants completed online psychological questionnaires at three time points: before the intervention (t1), after 3 months (t2), and after 6 months (t3). The intervention followed a crossover design: between t1 and t2, IG1 received the app-based resilience training, while IG2 did not. Between t2 and t3, IG2 received the resilience training, while IG1 did not. All questionnaires were administered at each measurement time, except for the BSI-18, which was only presented at t1, and the SUS, which was given to the group that had just completed the app-based resilience training (either at t2 or t3).

The online assessments took 20–30 min to complete, and the resilience training was suggested to be performed 10–60 min per day during the training period of 2 month. After the completion of the study, all participants were given the option to additionally participate in tablet-based training. Figure 2 shows a visual representation of the study process.

All analyses were performed using IBM SPSS software (Version 29; IBM Corp.). To simplify the analysis and increase sample size for more robust results, we created three key variables: Baseline: This included data from both IG1 and IG2 at t1. Post Intervention: This included data from IG1 at t2 and IG2 at t3. Post Control: This included data from IG2 at t2 and IG1 at t3. These variables were used to analyze the RS-13, PSQI, and SCI scores. This approach allowed us to combine data across different time points and groups, increasing the statistical power of our analyses. To better understand the impact of the resilience training on mentally vulnerable individuals, a median split for resilience (RS-13), psychological distress (BSI-18), and stress (SCI) was conducted and analysis was calculated for RS-13, PSQI, and SCI.

Descriptive data were generated for all variables. Moreover, bivariate Pearson correlation analyses were conducted to observe single relationships between the test variables. Variables were tested for normal distribution. For the purpose of this study, only the independent modules SCI Stress Score, SCI Somatic Score and SCI positive coping as well as SCI active coping were analyzed, as previously reported by [43]. To test our main hypotheses, repeated measurement analyses of co-variance were used to examine the effect of the resilience training on resilience (RS-13 score), and stress (SCI stress score and SCI somatic score) and coping (SCI positive coping and SCI active coping), across the three measurement times, controlling for sex [4], age, and education [10]. These models account for the possibility of carry-over effects and period effects [44]. Significant main- and interaction effects were followed up with Bonferroni-corrected pairwise t tests and Cohen’s d_z. The usability of the app was rated with the SUS as good (M = 77.06, SD = 17.30, [45]). All hypotheses were tested two-tailed at an α-level of 0.05. Assumptions to conduct the statistical tests were fulfilled unless otherwise noted.

Only individuals between the ages of 18–60 were included in the study procedure. Contrarily, participants were excluded if they had a lifetime diagnosis of a mental disorder (e.g., affective disorder, anxiety disorder, psychosis, dementia), continuous psychotherapy, or had cognitive or physical disabilities that would not allow the use of the resilience training (which was clarified during the inclusion interview). In total, 71 data sets were collected. The final data consisted of 34 participants (n = 8 male and n = 26 female; M (age) = 35.3 years, SD (age) = 9.3 years, range (age) = 24–59 years), including only complete data sets. Of this sample, 8.8% (n = 3) had a formal education, 20.6% (n = 7) had a high school diploma, 32.4% (n = 11) obtained a bachelor’s degree and 38.2% (n = 13) a master’s degree.

Results showed no significant main effect between baseline, post intervention and post control condition for the resilience score RS-13 (F(32, 2) = 1.30, p = 0.278, η_p² = 0.04), when controlled for age, sex, and education level. This indicates that resilience did not change across measurement times. Regarding the SCI stress score (F(34, 2) = 1.10, p = 0.368, η_p² = 0.03), the analyses of co-variance analyses showed no significant main effect, when controlled for sex and education level. However, when controlled for age, the SCI stress score showed a significant main effect (F(34, 2) = 3.73, p = 0.030, η_p² = 0.11; see Fig. 3). The SCI stress score at baseline was significantly higher than after the intervention (p = 0.036). However, there was no significant difference between the post-intervention and control group (p > 0.235), or baseline and control group (p > 0.999).

There were no significant effects for SCI somatic score (F(32, 2) = 1.065, p = 0.351 η_p² = 0.04), SCI positive coping score (F(34, 2) = 0.143, p = 0.867, η_p² = 0.01), and the SCI active coping score (F(34, 2) = 0.078, p = 0.925, η_p² = 0.00), when controlled for age, sex, and education level. These results show that neither somatic complaints nor coping styles changed across measurement times.

No statistically significant difference between the sexes was found. Means and standard deviations of the questionnaire scores at each measurement time are illustrated in Table 1.

To further investigate these findings, we analyzed the correlations between various scores, as shown in Table 2. We found a significant negative correlation between the SCI stress score after the intervention and the RS-13 score at baseline (r = −0.43, p = 0.011). This indicates that higher baseline resilience (RS-13) is associated with lower stress levels (SCIs) after the intervention. There were significant correlations between the SCI stress score and the BSI subscales at baseline: “somatization” (r = 0.43, p = 0.011), “anxiety” (r = 0.43, p = 0.012), and “depression” (r = 0.59, p < 0.001), as well as a significant positive correlation between the SCI stress score and the BSI Sum Score (≙ GSI, r = 0.56, p < 0.001) at baseline. This means higher general psychological distress (GSI) is associated with higher stress levels (SCIs). There was no evidence of a correlation between age and the SCI stress score (r = 0.22, p = 0.21), indicating that age does not significantly impact stress levels measured by the SCI.

A median split for resilience, SCI stress, GSI, BSI-18 somatization, BSI-18 depression and BSI-18 anxiety at baseline was calculated. Participants with low resilience at the beginning of the intervention showed lower scores across the measurement times, compared to participants with high resilience at the beginning (p = 0.002, η_p² = 0.27). Furthermore, participants achieved higher resilience scores in the post-intervention and post-control condition, when presenting with a low GSI score (p = 0.001, η_p² = 0.98), low BDI depression score (p = 0.002, η_p² = 0.26), or a low BDI anxiety score (p = 0.035, η_p² = 0.132). The opposite was identified for the SCI stress score: participants presenting with low resilience scores at T1, showed higher SCI stress scores across the measurement times (p = 0.021, η_p² = 0.16), and lower stress scores when presented with a low GSI score (p < 0.001, η_p² = 0.39), low BDI depression score (p = 0.006, η_p² = 0.26), or a low BDI anxiety score (p = 0.008, η_p² = 0.20).

The current study aimed to investigate the effect of using the newly developed AI-Refit app to improve resilience, stress regulation, and mental health in healthy individuals. The self-designed app, which was based on the concept of “7 pillars of resilience” [42], was used for 2 months (three measurement times). Compared to the control condition, there were no improvements on the resilience scale, but there were improvements in stress regulation measured with the SCI. These improvements are also clearly linked to mental health by significant correlation with the BSI, a screening procedure for mental disorders. Furthermore, there are no differences between men and women. Age plays a role and moderates the effect.

Resilience does not change as a result of the intervention. This contrasts with previous research programs like RESIST, which could show a medium to large-sized effect of resilience training [35]. This program, however, was built around the theoretical framework of the “Positive Appraisal Style Theory of Resilience.” Participants were asked to collect moments of resilience, develop a resilient self-image, and apply it to future challenges. In contrast, the framework of the “7 pillars of resilience” [42] was used for the present study, which has a broader focus that includes self-efficacy, optimism, and empathy, but also movement and cognitive training. To our knowledge, there are no published studies evaluating this specific resilience program, yet. Critics have pointed out that the uneven application of frameworks and different formats of resilience training poses significant challenges to the comparability and accurate measurement of results [46, 47].

Resilience is considered a crucial factor in managing stress [43] and lowering psychosomatic symptoms [44]. Numerous studies have found that resilience training is positively associated with stress, mental health, and well-being (e.g., [14, 45]) at both the psychological and biological level [48]. In contrast, a study on college students found no evidence that stress or somatization symptoms could explain HRV [49].

According to the post hoc results of this study, when controlled for age, stress regulation abilities improved during resilience training. This aligns with early research, indicating that older adults often display higher resilience due to the accumulation of life experiences and a greater ability to effectively manage stress and adversity [50].

A further study, utilizing an Israeli sample, demonstrated that stress levels, as measured with the BSI, significantly decreased and resilience levels significantly increased across three age groups, spanning from 18 to 91 years. Both variables were found to be significantly correlated with each other [51].

Current studies indicate that the initial indications of mental illness can be demonstrated through somatic complaints [52] and Smith et al. [53] found a linear relationship between the improvements in resilience and the decrease in reported stress and somatic symptoms. This aligns with our results demonstrating a correlation between resilience and stress, as well as resilience and psychosomatic symptoms. The App and the program used in this study might have fostered this connection since the main features included psychoeducation, self-reflection, mindfulness, and relaxation, which have also proven to lower stress and psychosomatic symptoms [52, 54]. Even though this was not directly addressed in the pillars of this program, it should be implemented to enhance the effects of stress regulation and coping. Enhancing stress regulation might not have a direct effect on improving resilience, given that resilience is a multifaceted construct, also influenced by environmental factors and genetic predispositions [1, 55].

Furthermore, correlations between stress and mental health, measured with the BDI, were found. Charles et al. [56], propose that even seemingly minor affective daily stressors can significantly influence mental health outcomes. It is therefore likely that resilience programs and stress reduction programs may strengthen mental health in nonclinical samples and as a consequence, function as an early subclinical prevention strategy.

Moreover, a training program might enhance mental health symptoms without improving overall psychological resilience. The RS-13 may not have been sufficiently sensitive to detect potential changes [8], as the scale measures items such as: “When I have plans, I follow them through.” which are considered more general, compared to the SCI, where items measure the current stress load within the last 3 months, with items like: “To what extent have you felt overwhelmed by the following events and problems in the last 3 months?”. Compared to stress, which focuses on every-day-stressors, resilience is more described as the ability to cope with significant life adversities. Therefore, the evaluation of the efficacy of a resilience training should be based on the response to adverse life events [57]. Including multiple measurements of resilience might allow further insight to understand the impact of the program on specific factettes of that resilience program. This data can be used to develop more precise resilience programs.

However, with a crossover design as presented in this study, possible carry-over effects and period effects might impact the results. Possible nocebo and placebo effects regarding the control condition must also be taken into account, however, were not included in the statistical analysis. Although a baseline condition was included, results must be interpreted cautiously [58]. Since participants emerged in the program at home, it was not possible to consider the potential stressors that they might have been facings or to verify the actual implementation of the resilience exercises. Consequently, only a self-report is available for verification. Furthermore, a larger sample size and additional demographic data, such as marital or socioeconomic status, may have provided beneficial insights.

The level of satisfaction “I think I would use the app regularly,” 1 = “I totally disagree,” 4 = “I totally agree” with the app as measured by the System Usability Scale (SUS) was relatively low (M = 1.64, SD = 1.05). The incorporation of a playful character within the app may enhance the frequency of its utilization. With the rise of online programs, the implementation in a web application, rather than an app, might even bring more promising results, as stated by Ang et al. [46]. Nevertheless, digital resilience programs conducted at home can yield promising results, mainly due to the flexibility of use. To further enhance the experience and raise user satisfactions, it would be beneficial to include more gamification and implement this intervention in a virtual reality environment.

The resilience intervention demonstrated an indirect effect on stress regulation and psychosomatic complaints in a healthy population, and it is possible that the program may prove more beneficial for those with mental health issues. In future investigations, it might be possible to analyze the effects on a vulnerable population or those with mental disorders, to gain a deeper understanding of the link between resilience and mental disorders and help a greater number of people.

A newly developed 2-month app-based resilience intervention, based on the seven pillars of resilience [42] was tested to improve several parameters of mental health in a healthy population. While there was no direct effect of the resilience intervention on the parameter resilience, there was an indirect effect on stress regulation and psychosomatic symptoms. These preliminary findings suggest that the present resilience intervention AI-Refit might be a valuable tool for enhancing mental health and regulating stress on multiple levels. Future randomized controlled trials in mentally ill and vulnerable populations will follow.

All participants gave written informed consent before they participated in the study, which was conducted under the Declaration of Helsinki and was approved by the Ethics Committee of the Medical University of Graz (EC-No. EK 49-297 ex 22/23).

Prof. Eva Z. Reininghaus and Prof. Dr. Nina Dalkner were members of the journal’s Editorial Board at the time of submission. The authors have no other conflicts of interest to declare.

This research has received funding from the FFG Project (No. 887487). This work was supported by the JOANNEUM RESEARCH Forschungsgesellschaft GmbH (2022–2024). The study design was approved by the funder in the application process. The funder had no role in the design, data collection, data analysis, and reporting of this study.

Melanie Lenger, Elena M.D. Schönthaler, Eva Z. Reininghaus, and Alina Hantke: manuscript conception, writing – original draft, and writing – review and editing; Melanie Lenger, Elena M.D. Schönthaler, and Alina Hantke: data analysis; Melanie Lenger, Elena M.D. Schönthaler, Suher Guggemos, Martin Pszeida, Jochen A. Mosbacher, Sandra Draxler, Thomas Lutz, and Jama Nateqi: conceptualization, study design, and execution; Nina Dalkner, Silvia Russegger, Jama Nateqi, Dietrich Albert, Lucas Paletta, and Eva Z. Reininghaus: supervision and writing – review and editing.

The data that support the findings of this study are not publicly available for data protection reasons due to the small number of participants, but can be requested from the corresponding author E.M.D.S ([email protected]).