The effects of computer-based mindfulness training on Self-control and Mindfulness within Ambulatorily assessed network Systems across Health-related domains in a healthy student population (SMASH): study protocol for a randomized controlled trial

The primary objective of this trial is to determine whether practicing momentary mindfulness (first primary outcome) mediates momentary self-control in daily life (second primary outcome). Of further interest are changes in habitual mindfulness and self-control (secondary outcomes) in the intervention condition compared to a wait-list control condition. Further outcomes are momentary attention control, affect and emotion regulation and key behavioral domains influenced by self-control (eating behavior, sexual behavior, work behavior, interpersonal behaviors, alcohol consumption, spending money).

A novel network approach [24] will be used to model the interplay of momentary self-control with internal and external factors, such as affect, desire strength and positive or negative events, over time. These time-to-time relationships will be analyzed at individual and group level by using multilevel VAR models. Afterwards, different network parameters will be computed exploring overall and local network structure (see “Statistical analysis”). In doing this, (1) underlying processes of self-control enactment become visible, (2) predictors of self-control from a range of different domains will be identified and (3) it is possible to examine mediating processes that convey the effect of mindfulness training on momentary mindfulness, and ultimately on self-control in daily life.

The first primary outcome of this trial is momentary mindfulness assessed via electronic diaries six times per day on 40 consecutive days between 10 a.m. and 8 p.m. with three selected items from the state version of the Mindful Attention Awareness Scale (MAAS; study 4 from Brown and Ryan [27]). The state version of the MAAS consists of five rephrased items drawn from the original MAAS (Items 3, 8, 10 13 and 14) in order to assess the degree of mindfulness experienced directly in daily life [27]. We chose only three items from the German MAAS [28] (Items 8, 10 and 14) to lower participants’ burden and adapting to the needs of an Ambulatory Assessment. The three items were rephrased according to Brown and Ryan [27].

The second primary outcome is momentary self-control which is measured via electronic diaries six times per day on 40 consecutive days between 10 a.m. and 8 p.m. with three items from the German version State Self-control Capacity Scale (SSCCS) [29]. The SSCCS consists of 25 items and assesses self-control capacity that is momentarily available for self-control execution (Cronbach’s α ≥ .93). For the SMASH study, the original SSCCS items 12, 16 and 18 were selected due to their high loadings on the SSCCS: “I can’t absorb any information,” “I want to give up,” “I feel like my willpower is gone.” The selected SSCCS items are complemented with items that were used by Hofmann et al. [7] (see Additional file 2: Table S1 and Table S2).

One of the secondary outcomes is habitual self-control which is measured with the brief Self-control Scale (SCS) consisting of 13 items [2]. The SMASH study uses the German Self-Control Scale-K-D (SCS-K-D) (Cronbach’s α = .79 (t1) and α = .80 (t2) and a test-retest reliability of r = .82) [30]. The SCS-K-D is answered at seven weekly meetings in order to measure changes in habitual self-control.

As a further secondary outcome, habitual mindfulness is assessed with the original version of the MAAS [27], which consists of 15 items. A validated German version of the MAAS is used in this study (Cronbach’s α = .83; test-retest reliability r = .82) [28]. The self-report of habitual mindfulness is collected at the seven weekly laboratory meetings.

Participants are asked to report their (1) age, (2) gender, (3) highest graduation, (4) engagement in a religion, (5) Yoga practice and (6) substance use (alcohol, cigarettes, marijuana).

The frequency of experienced desires in daily life is measured with 15 desire domains [7, 8].

Also goal commitment is assessed by asking whether participants have planned to pursue 20 specific goals [7, 8] for the next 6 weeks. Personality traits are assessed with the German version of the Big Five Inventory-44 (BFI-44) [31, 32], which consists of 42 items and five scales (Cronbach’s α = .63 to α = .86; test-retest correlation r = .55 to r = .82) [32]. Furthermore, social desirability is assessed with the Social Desirability Scale-17 (SDS-17) [33, 34] (Cronbach’s α = .72; test-retest correlation r = .80) [34] consisting of 16 items.

Complementary to the MAAS, different aspects of mindfulness (observing, acting with awareness, acting without judgement, and describing) are assessed with the 39 items from the German Kentucky Inventory of Mindfulness Skills (KIMS) [35, 36] (Cronbach’s α = .79 to α = .92; test-retest correlation r = .78 to r = .86) [36]. Additionally, two factors of mindful experience (curiosity and decentering) are measured with the Toronto Mindfulness Scale (TMS) [37], which consists of 13 items (Cronbach’s α = .95).

Wellbeing is measured with the German version of the WHO-5 Wellbeing Index [38] consisting of five items (Cronbach’s α = .92 [39]). As another measure for psychological wellbeing, the German version [40] of the Brief Symptom Inventory-18 [41] is additionally used to assess changes in syndromes of somatization, depression and anxiety with 18 items (Cronbach’s α ranges between α = .63 and α = .93) [40].

Moreover, habitual behaviors in several self-control domains are assessed. The use of two different emotion regulation strategies, reappraisal and suppression, is measured with the German version of the Emotion Regulation Questionnaire (ERQ) [42, 43] which consists of 10 items (reappraisal α = .76; suppression α = .74) [43]. Interpersonal relationships are measured with the German version of the Adult Attachment Scale (AAS) [44, 45] consisting of 15 items and the three scales of closeness, trust and fear (α = .72 to .79) [45]. The habitual eating styles restrictive eating, emotional eating and external eating are assessed with the Fragebogen zum Ernährungsverhalten-II (FEV-II, Cronbach’s α = .82 to α = .92) [46], the German version of the Dutch Eating Behavior Questionnaire (DEBQ) [47]. The FEV-II consists of 33 items.

In order to investigate affective dynamics, the measurement of affect is based on Russell’s affective circumplex model [48, 49]. Positive and negative affect are each assessed with four items.

As a behavioral measure of mindfulness, the Breath-counting Task is conducted [26]. During this task, participants have to count their breath for 18 min repetitively from one to nine. Additionally, participants are asked to press a button whenever they caught themselves miscounting their breath (self-caught miscouting). Counting accuracy will be analyzed by calculating the percentage of correctly counted breath trials in relation to all trials (self-caught trials, miscounted trials and correctly counted trials). The breath counting is interrupted every 90 s on average in order to present two items, assessing momentary meta-awareness and mind wandering [26].

Due to conditional branching in the presentation of items during the Ambulatory Assessment, participants answer 31 to 50 items (Additional file 2: Tables S1, S2 and S3) at each signal six times per day regarding (1) positive and negative affect [48, 49], (2) worrying [24, 50], (3) momentary emotion regulation strategies [12], (4) attention (item from previous research in our laboratory) and (5) situational circumstances (items from several Ambulatory Assessment prompts [8, 11, 12, 51]). If a participant indicates a specific desire, additional items appear on the smartphone screen depending on the indicated desire: for example, if a given participant indicates the desire “food”, items are presented that assess specific eating behaviors. Further additional items from the (6) DEBQ [47], (7) the Situational Motivation Scale [52], (8) the Work-related Flow Inventory [53], (9) the State Adult Attachment Measure [54], (10) the Recovery Experience Questionnaire [55] and items regarding (11) sexual behavior and (12) spending money are presented in Additional file 2: Tables S2 and S3. A pre-test (N = 6) showed that it takes about 2 min (M = 1.89; SD = 1.84) to complete the presented questionnaires after each signal.

At the time point of study enrollment, there was no method that could model different types of variables in a network. Due to the chosen mode of statistical analysis (network approach) that includes VAR and multilevel regression models, ambulatorily assessed variables all had to be either continuous or binary variables. For this reason, all originally binary variables were changed into continuous variables.

During the first laboratory session, all participants complete self-report questionnaires and a Breath-counting Task [26] which is conducted on computer. One week after the first laboratory meeting, both intervention and control condition participants come to five weekly laboratory meetings completing questionnaires regarding self-reported dispositional mindfulness and self-control. A maximum of three participants can take part in a laboratory meeting at once. From the second laboratory meeting on, only the intervention condition starts practicing mindfulness with a computer-based guided-breathing meditation for 12 min at five weekly appointments. At the seventh and final laboratory meeting, a post-intervention measurement takes place. During the final measurement, questionnaires that were completed within the first laboratory meeting are answered again. Also the Breath-counting Task [26] is conducted by both control and intervention conditions comparing counting accuracy between them as well as before and after mindfulness training. The seven laboratory meetings are conducted on computers with the operating system OS X El Capitan, Version 10.11 and a screen resolution of 1920 × 1080. The software Inquisit 4 (Millisecond Software, 2014) is used to present the questionnaires on the screen.

A day after the first laboratory session, the Ambulatory Assessment begins (day 1 till day 40). Each day, signals are randomly distributed six times a day via smartphones throughout a defined timeframe of 10 h. Signals can occur between 10 a.m. and 8 p.m. This time span is divided into blocks of 2 h [56], defining for each block that two consecutive signals could be at least 45 min apart. Throughout a timeframe of 5 min after a signal, participants can start, reject or postpone their answer for 15 min. Participants are asked to complete short questionnaires which assess experiences at the present moment or within the last 30 min [7].

Moto E (second generation) smartphones with the latest Android Lollipop version 5.1 are handed to participants in case they do not own an Android-based smartphone. Ambulatory Assessment is run via the application movisensXS, version 0.8.4203 (movisens GmbH, Karlsruhe, Germany), which has to be downloaded and installed either on the personal or the received smartphone. This app schedules the beeping signals, presents the questionnaires on the smartphone and saves collected data by uploading it to a secured server.

Dynamic relations between self-control variables will be analyzed and visualized with the network approach [24] for the baseline (week 1) and post-intervention period (week 6) as well as for other timeframes. For computing estimates of the multilevel VAR models, the R package mlVAR will be used [61]. Networks will be constructed out of selected ambulatorily assessed variables (e.g., self-control depletion, mindfulness, negative affect, negative event, conflict and desire strength). First, multilevel VAR models are modelled for each network variable – every single network variable is regressed to a lagged version of itself and all other independent lagged variables (values that were measured at a previous time point). These multilevel VAR models contain fixed and random effects where the fixed effects are the average population effects of the lagged variables on the dependent variable (group level) and the random effects the participant’s deviations of the average population effect (individual level). Then, the fixed-effects coefficients of all modelled multilevel VAR models are used to visualize directed relationships between all network variables. The self-control population network containing average connection strength between variables at group level will be visualized with the qgraph R package [62]. In accordance with Bringmann et al. [24], the False Discovery Rate [63] will be controlled at 5% in order to control for multiple testing. An exemplary network is presented in Fig. 2 showing time-to-time relations (edges) between network variables (nodes). For a more detailed description, see Bringmann et al. [24]. This procedure will also be applied to construct population networks for the control as well as for the mindfulness condition in specific health domains like alcohol consumption, emotion regulation and eating behavior.

By using a subject’s individual random effects with the fixed-effect coefficients of the multilevel VAR models, networks can be visualized that show only temporal dynamics of a specific subject. In this way, individual self-control networks can be visualized [24].

In order to investigate the effect of mindfulness on the interconnections of the self-control network, a treatment variable (mindfulness versus control) and a time period variable (pre versus post intervention) will be added to the multilevel VAR models of the population network [24]. Differences between mindfulness and control conditions can be examined by looking at single links between two specific variables within the network [24]. If there is a change across time at a specific link between a lagged independent and a dependent variable that differs between conditions, there should be a significant three-way interaction between the variables treatment, time period and the independent lagged variable on the dependent variable.

To assess local and overall network structures and their changes, several network parameters, such as network density (overall connectivity) and centrality (In-strength, Out-strength, Closeness and Betweenness), will be analyzed for individual and population networks [64]. Density indicates the overall connection strength between all nodes (variables) and is calculated by averaging over the absolute values of all regression coefficients. High density represents strong interconnections between the network variables which means they have high impact on each other from one time point to the next. Different centrality indices [65] will be computed for all network variables to explore the importance within the network of each of them. Those indices indicate how much information a specific node exchanges with other nodes or how close it lies to the others (for a detailed description see Bringmann et al. [24, 64]).