This pilot study used a Single Case Experimental Design (SCED). In a SCED all participants provide data multiple times, both before and after an intervention starts, so that they become their own control group (34). In addition, to avoid bias by possible unrelated events (history bias), all respondents started at different moments. An ABA-structure was used, meaning that one intervention phase was included (phase B) and surrounded by two phases (A) in which use of the SAM Junior app was prohibited. Data was obtained at four moments separated by four weeks: Control, pre-test, post-test and follow-up (see Figure 1 ). Based on Hoeberichts et al. (30) study on the SAM for adults, it was assumed that this time window would be enough to measure potential effects. Outcomes were perceived stress, adaptive and maladaptive coping styles and QoL.

Participants were recruited at an outpatient department of two Dutch mental health care facilities, GGz Centraal and Karakter. Both departments provide specialized care for autistic adolescents. Inclusion criteria were (a) a diagnosis of autism according to DSM-V or DSM-V-TR, given by a licensed professional, (b) sufficient intellectual skills and willingness to use the SAM Junior properly, and (c) an age between 12 and 18 years old. Participants were recruited by their mental health professionals as well as through recruitment posters and flyers that were spread throughout the aforementioned organizations. During the course of this study, participants continued to receive care-as-usual.

Based on a power analysis of 1600 Monte Carlo-simulations with a medium effect size of d = 0.5, α = .05, and power = .80, 19 participants were needed to reliably answer the research questions. In order to meet this requirement, and to account for potential drop-out, a total of 24 participants were included.

SAM was originally developed by GGz Centraal in collaboration with the Dutch organization of applied science research (TNO). Several changes were made to the junior version, such as the addition of age-appropriate activities (e.g. ‘doing homework’, ‘exams’) and stress tips (e.g. ‘watch a movie on your smartphone’), a stress tip appearing immediately upon opening the app, and a longer window of two hours to complete the questionnaires. The changes were based on a co-creation phase with adolescent autistic clients of GGz Centraal. Some personalization options in the SAM Junior app are available (e.g. written text or emoticons in the questionnaire, color schemes and the ability to add personalized activities). No supervision is needed to use the app. The app can be downloaded from the Google Play Store and App Store in 33 countries around the world in seven languages at no cost. More information about the SAM Junior app and the development can be found on https://www.stressautismmate.nl/samjunior/.

All participants started with a face-to-face or online appointment with a researcher of this study. During this meeting, the first set of questionnaire data was collected through an interview with the researcher. In addition, informed consent and demographic data (age and gender) were obtained. After questionnaire data were obtained during the second (pre-test) appointment, participants installed the SAM Junior app on their mobile phone and were guided through the app and its settings by the researcher. Throughout the intervention phase, a helpdesk was available for all questions and technical issues. During the third (post-test) appointment, participants de-installed the app and were told not to use the app until the follow-up phase had finished. After the participant’s follow-up data were collected, they were free to use the app again.

Linear mixed-effects models were used in SPSS (version 28.0.1.0), which accounts for the fact that effects may differ between individuals (42). Since the participants in this study acted as their own control group, this type of model was suitable. Time of measurement was a fixed (between-subjects) factor, while the participant was a random (within-subject) factor. Participants who did not complete all phases were excluded from the data analysis. Shapiro-Wilk (PSS: W(64) = .99, p = .91; CERQ-Adaptive: W(64) = .99, p = .76; CERQ-Maladaptive: W(64) = .98, p = .33; KIDSCREEN-27: W(64) = .97, p = .07), Durbin-Watson (PSS: d = 1.49; CERQ-Adaptive: d = 2.31; CERQ-Maladaptive: d = 1.87; kidscreen-27: d = 1.53) and visual homoscedasticity tests showed that the data met the assumptions of normality, no significant autocorrelation and homoscedasticity. Therefore, no corrections in order to build the linear mixed-effects models were needed.

For each individual model, it was calculated whether effects occurred on the sum scores of the PSS (perceived stress), CERQ (adaptive and maladaptive coping styles) and KIDSCREEN-27 (quality of life between the pre-test (t1) and post-test (t2) phase, as well as the pre-test (t1) and follow-up (t3) phase. An alpha of.05 was used. Exploratory analyses on the subscales of the CERQ and KIDSCREEN-27 were also conducted. In these analyses, due to the large (28) amount of tests, the Holm-Bonferroni correction for multiple testing was used.

As we did not document participants’ user experiences in this study, we can only speculate on the lack of effects of the SAM Junior app. First, the app may have failed to reduce stress levels because it was too complex for the study population. The use of mHealth requires several executive functions, such as initiating use and maintaining attention, as well as a certain amount of self-understanding (43). These abilities are generally less well developed in autistic adolescents (44, 45). Compromised executive functioning and a heightened rigidity in adolescents with autism may also have caused difficulty in applying the stress tips (46, 47), especially considering the relatively short time span of four weeks in which the app was used. Executive function problems and cognitive inflexibility may also have caused the high amount of dropout during the intervention phase. As executive functions still develop in early adulthood, (44), it may explain why autistic adults in previous studies (30, 31) were able to benefit from the SAM app, while the SAM Junior app showed no effects in autistic adolescents.

Alternatively, the use of SAM Junior may have caused extra stress in users, compromising its potential effect (48). Technostress caused by mHealth has been widely documented (49). Having to fill in a questionnaire at several fixed moments per day may have caused users to change their daily schedule, compromising their autonomy (50). For example, as mobile phones are often banned from classrooms in the Netherlands, participants may have felt compelled to use their limited break time to complete questionnaires. This indicates that the app may not have been tailored well enough to their daily schedules. Furthermore, participants’ heightened stress scores could have reminded them of their low well-being, causing more stress as a result (51, 52).

Finally, the SAM Junior app may have been ineffective due to the study population of clinical outpatients. These participants may have had higher support needs and extra stress caused by common co-occurring conditions such as anxiety disorders and post-traumatic stress disorder (53), meaning aid in managing general daily stressors was insufficient. Moreover, crisis prevention plans, that are recommended in Dutch clinical autism practice (54), serve a similar function as the SAM Junior app, potentially making the app redundant. That said, the SAM app for adults was effective in a clinical population (30, 31), meaning this alone does not fully explain the null results.

There were no changes observed in maladaptive and adaptive coping styles after use of the SAM Junior app. The stress tips of the SAM Junior were predominantly focused on behavioral change, while the CERQ measures cognitive coping strategies. This may explain the lack of change in CERQ scores. Regardless, some stress tips, such as do a breathing exercise, have previously shown to alter both adaptive and maladaptive cognitive coping styles in autistic adults (55). Therefore, the lack of effects of SAM Junior on coping styles suggests that some useful stress tips may not have been chosen as often or were not used as intended.

Although not statistically significant, the average QoL in participants dropped during the intervention phase. Regression to the mean may be responsible for this pattern (56). Indeed, average quality of life at baseline was higher than is commonly found in adolescents with autism (20, 57–59) and also higher compared to a cohort of Dutch allistic adolescents (60). Furthermore, it has to be noted that a change in QoL was less plausible than a change in the other variables, as perceived stress is only partially linked to QoL (61–63). This also explains why the perceived stress levels and QoL of the participants could be high at the same time.

This study has several strengths. This is the first quantitative study that evaluated an mHealth application that aims to reduce stress in adolescents with autism. The app was created in co-creation with autistic adolescents which assured that the app was tailored to their needs. The questionnaires were conducted during a meeting with a researcher, which enabled the participants to ask the researcher about any unclear questions.

Several limitations also need to be mentioned. First, partially due to drop-out, this study was underpowered. This means we were not able to detect potential smaller effects. As the SAM Junior app is free and available to anyone with a smartphone, even small benefits to the well-being of autistic adolescents would have been a relevant finding. Second, although we aimed to gather data strictly at four-week intervals, some participants took longer to respond, often due to poor mental health or scheduling conflicts. This may have changed outcomes, as participants might have only completed the questionnaires during lower stress periods or may have used the app for longer than intended. Third, the adherence of the participants was not checked, meaning it is unsure to what extent the app was actually used. Therefore, a potential lack of use of the SAM Junior app may also explain the null results of this study. Fourth, due to the small sample size and the fact that all participants received assistance from a mental health care facility, as well as the exclusion of autistic adolescents with intellectual disabilities, the study sample may not have been representative of the all adolescents with autism.

Qualitative research about the user experience of the SAM Junior app may help to gain a better understanding on why the app was not found to be effective. We are currently conducting such a study. Based on its outcomes, the app can be improved. Additionally, recommendations from existing literature can also guide enhancements. Here, we will provide three examples. First, implementing text-to-speech options could better address the unique sensory needs of adolescents with autism (64). Second, adding gamification features (e.g. rewards for lowered stress levels) may increase user engagement and motivate users to apply stress tips in daily life (65). Third, allowing users to complete the in-app questionnaire at their preferred times, instead of at set intervals, may reduce the stress associated with using the app (50).

Once the SAM Junior app has been upgraded, the new version can be tested in a quantitative study. To increase the certainty that potential effects are caused by the use of the app, a SCED with an increased number of phases (i.e. an ABAB-design) is recommended. Alternatively, to increase the validity of the outcomes further, a randomized controlled trial can be conducted. In addition, collecting data at more time points would allow for mediation analyses. These can help to discover the exact mechanisms through which the SAM Junior app may be effective (66). Regardless of the research design, we recommend studying a non-clinical sample of autistic adolescents to prevent potential confounding effects of concurrent treatment on the study outcomes.

In sum, the results of this study do not show effectiveness of the SAM Junior app in adolescents with autism. This implies that the app may not currently be a viable alternative for reducing stress in this population. However, the limitations of this pilot study warrant caution in interpreting its outcomes. Further research with an improved application is recommended to determine the potential effects of the SAM Junior app with more certainty.