1.1. Participants
Participants were recruited in March 2019, via referral from a mental health community centre in the French speaking community of Belgium, where the recruitment, assessments and intervention took place. Inclusion criteria for the present study were: aged between 18 and 65, met DSM-5 criteria for schizophrenia or schizoaffective disorder [24] and a good understanding of French. Exclusion criteria were: presented an unstable clinical picture (i.e., no acute positive symptoms); evidence of a significant change in medication within one month prior to baseline assessment; history of severe brain trauma or epilepsy; comorbid intellectual disability; and moderate or severe substance use disorder other than tobacco (according to the DSM-5; i.e., showing 4 or more symptoms). The head psychiatrist from the mental health community centre was familiar with the inclusion and exclusion criteria of the study and other relevant details (e.g., the need for participants to provide informed consent). Thereafter, out of a pool of 60 patients, six candidates who fulfilled the criteria were contacted, introduced to the study and asked if they accepted to be contacted by the main investigator. Three accepted. Next, the main investigator contacted these three candidates by phone and presented the study. The participants were then seen in person and received a thorough explanation of the evaluation protocol, the intervention and their rights as participants in the study. They were invited to read the information sheet (including repetitive disclosure and emphasis of key points, as recommended by [25]), ask any questions that they might have, and sign the informed consent if they accepted to participate. Three participants enrolled in the study and were assigned to the Switch intervention in April 2019. One participant was lost to follow-up in October 2019.
The study was approved by the Liege University Hospital Ethics Committee (B707201629105). Sociodemographic and clinical characteristics of the participants are reported in Table 1.
[INSERT TABLE 1 HERE]
1.2. Study design and procedure
Participants underwent three types of evaluation: traditional assessment scales of motivational deficits, apathy, quality of life and daily functioning (pre, post and follow-up); ambulatory assessment including ESM (i.e., questionnaires); actigraphy (step count) (see Figure 2). Participants received feedback on all evaluations at the end of the study.
[INSERT FIGURE 2 HERE]
Figure 2. Design of the study and assessment procedure. BNSS = Brief Negative Symptom Scale; LARS = Lille Apathy Rating Scale patient and informant; FROGS = Functional Remission of General Schizophrenia, S-QoL = Schizophrenia Quality of Life questionnaire.
1.3. Traditional evaluation (primary outcomes)
Assessments were completed by trained evaluators. Participants were told to refer to the last 2 weeks when completing the following scales:
Brief Negative Symptoms Scale (BNSS) [26]. The French version of the BNSS was used in this study [27]. The BNSS assesses expressive and motivational negative symptoms. Only the BNSS – Motivation subscale was used, which is the mean of the following subscales: anhedonia (intensity of pleasure during activities, frequency of pleasure, intensity of expected pleasure from future activities), asociality (behaviour, internal experience), avolition (behaviour, internal experience). Each item is scored from 0 to 6 (0 = no impairment; 1 = very slight; 2 = mild; 3 = moderate; 4 = moderately severe; 5 = marked; 6 = severe). A blinding procedure was used: participants’ interviews were recorded and each video or sound recording was evaluated at the end of the study by two independent evaluators who were unaware of the recordings’ assessment time. The BNSS possesses excellent internal, convergent and discriminant validity [27], excellent test-retest and interrater reliability [26], and good sensitivity to change [28].
Lille Apathy Rating Scale – Patient version (LARS-p) [29]. The LARS is a semi-structured interview that evaluates the different dimensions (cognitive, emotional and behavioural) of apathy through the following subscales: everyday productivity, interests, taking initiatives, novelty seeking, voluntary actions, emotional responses, concern, social life and self-awareness. The total score ranges from -36 to 36 ([-36; -22] = absence of apathy; [-21; -17] = tendency towards apathy; [-16; -10] = moderate apathy; [-9; 36] = severe apathy). The LARS-p possesses a high level of inter-rater reliability and satisfactory internal consistency [30].
Schizophrenia - Quality of Life questionnaire (S-QoL) [31]. The S-QoL is a 41-item questionnaire that evaluates life satisfaction regarding psychological wellbeing, self-esteem, family relationships, relationships with friends, resilience, physical wellbeing, autonomy and sentimental life. Items are rated on a 5-point Likert scale (1 = much less satisfied than expected; 2 = less satisfied; 3 = slightly less satisfied; 4 = as satisfied; 5 = more satisfied). The total score ranges from 0 to 100, higher scores indicating better subjective quality of life. The S-QoL shows good internal and convergent validity, excellent test-retest reliability [31], and good sensitivity to change [32].
Informants were interviewed to provide an external understanding regarding participants functioning. The informant for participant 2i-1 was her husband; the informant for participant 2i-2 was the head of his supervised housing; participant 2i-4 did not wish to involve an informant. Informants completed the following two scales:
Lille Apathy Rating Scale – Informant version (LARS-i) [33] (see the patient version for a description). The LARS-i shows high internal consistency and concurrent validity, as well as high levels of test-retest and inter-rater reliability [33].
Functional Remission of General Schizophrenia (FROGS) [34]. The FROGS is a measure of daily life outcomes, which evaluates level of functioning in 5 different domains: daily life, activities, relationships, quality of adaptation, and health and treatment. Via a semi-structured interview with the informant, each item is assessed on a 5-point scale: 1 = does not do; 2 = does partially; 3 = does a significant part; 4 = does almost all of it; 5 = does perfectly. The total score ranges from 19 to 95. The threshold score for remission is 61 [35]. The FROGS possesses high concurrent validity and internal consistency [34].
1.4. Ambulatory assessment (secondary outcomes)
ESM questionnaire
Prior to the start of the study, the participants received extensive explanations regarding the ESM procedure. Participants installed the MetricWire app (https://metricwire.com/) and were logged in with a sham email address. Participants filled in an example-questionnaire with the investigator who explained all the questions and their possible answers.
For the 14 consecutive days of the baseline phase (see Figure 2), participants were prompted by the app MetricWire five times a day at pseudo-random time points, within 3-hour time frames between 7.30 a.m. and 10.30 p.m. Each prompt invited the participant to open the app and answer the questionnaire referring to what he/she was experiencing just before the prompt. The participants had 20 minutes to fill in the questionnaire and they received a reminder after 10 and 15 minutes. During the two months of intervention, in order to reduce the burden on the participants, the number of prompts was reduced to three per day, within 5-hour time frames. After the end of the intervention, participants were prompted again 5 times per day for another 2 weeks. As participants did not complete enough questionnaires after the end of the intervention, the post-assessment ESM observations (T2) were not taken into consideration.
The ESM questionnaire was developed based on the different variables included in the motivation model described in the introduction. It was created following guidelines from Kimhy et al.[36]. The questionnaire included 14 questions, plus three optional branched questions, i.e., determined by the participant’s answer to a previous question (see Table 2).
[INSERT TABLE 2 HERE]
The categorical ESM outcomes of interest were: activity’s meaning, motivation, mood, confidence, and savouring. Activity’s meaning, effort, energy, mood, and confidence represent each a single item from the ESM questionnaire (see Table 2). Motivation is a composite measure of the items motivation and wanting to give up (reverse coded). Savouring is a composite measure of present enjoyment, reminiscence, and projection into the future.
Nominal ESM outcomes of interest were coping strategies (in the presence of discouraging beliefs), social contact, activity and initiation. For further details and label descriptions of the ESM measures, please refer to Table 2.
Step count
Participants were provided with an activity band (MiBand 3, Xiaomi) which they had to wear at all times (day and night) during the different phases of the study (baseline, intervention, post-measurement, and follow-up). The band is waterproof and has a battery autonomy of approximately 20 days. A MiFit sham account was created in order to synchronize the activity band with the app on the participant’s smartphone. The MiFit app provided the total amount of steps per day.
1.5. Intervention
Switch was delivered by the main investigator, a trained psychologist and psychotherapist. The individual sessions lasted 1 hour and were given twice per week for 2 months, in the participants’ local mental health centre.
The first sessions were dedicated to building a therapeutic alliance, getting to know the person and identifying personal resources, goals and values (i.e., addressing the first two levels of the motivation model, see 1. Introduction). Strategies were then taught in order to help the person to engage in behaviours directed towards these chosen goals and values (i.e., moving to the third level of the model). Multisensory “imagery” was used to help to look forward to the future (i.e., pleasure anticipation). This type of projection into future actions/goals included not only visualising the scene (i.e., the person her/himself, the context, the actions), but also imagining the possible sounds, physical sensations, smells, flavours, pleasant emotions, constructive thoughts, etc. The imagery thus goes through the different senses, in order to increase the possibility of experiencing pleasure and to help identify what modality generates the more pleasure – and consequently that has the higher motivational power. The imagery could focus on the process (e.g., baking a cake) or the result (e.g., eating the cake). A restructured decisional balance tool was used to address the effort-value computation. The “motivation’s switch” (see Figure 3) was used to identify all the reasons why the person would not engage in a certain activity (including potentially discouraging thoughts, required effort), and all the good reasons why she/he would engage in that activity. Additionally, a column was used to indicate quick solutions for the smaller obstacles that were identified. The solutions and pros represent the “ON” part of the switch, which is highlighted relative to the cons’ column, which represents the “OFF” part of the switch. The cons column potentially included obstacles and dysfunctional attitudes that needed further attention. Participants were then guided in solution-seeking strategies. Furthermore, significant discouraging thoughts and low self-efficacy were challenged using cognitive restructuring (e.g., generating more constructive thoughts) and/or a cognitive defusion approach (e.g., using metaphors, training mindfulness). Help in planning and initiation strategies (electronic reminders, implementation intention, post-its…) addressed the subsequent steps in the model. Finally, participants were invited to use various reminiscence strategies (e.g., sharing of experience with others, keeping a diary, looking at photos, buying souvenirs) to increase positive memories and boost motivation for new actions or goals to engage in.
[INSERT FIGURE 3 HERE]
Figure 3. The motivation’s switch. Updated version of the decisional balance.
Each participant learnt the different strategies in relation to their individual goals and needs. A folder which presented the rationale for each strategy was given to the participants. Take-home assignments were given and stored in the same folder. Participants were also given small cards (that could fit in their wallet) containing the key elements of each strategy. Furthermore, during the last three weeks of the intervention, the participants received daily triggers (via the MetricWire app) inviting them to look forward to coming events/activities (morning trigger) and to look back at their day and reminisce about positive incidents (evening trigger). The morning trigger included an mp3 that could be listened to from the app and that provided a guided multi-sensory projection into the future. Finally, a booster session took place around 45 days after the end of the intervention, consisting mainly of a reminder of the different strategies and a troubleshooting of possible obstacles.
The complete manual for the Switch intervention and the participant booklet (both in French) can be requested from the first author.
1.6. Analyses
Aim 1: treatment effects of Switch on motivation and related processes
We used effect size coefficients to report changes in BNSS, LARS-p, LARS-i, FROGS and S-QoL scores. We created effect size coefficients for the composite scores of motivation/apathy and of QoL/functioning. Motivation/apathy represented the mean of four scores from the BNSS-Motivation (i.e., two scores from the two blinded evaluators), the LARS-p and LARS-i. QoL/functioning included the two scores from the S-QoL and the FROGS. First, all variables were rescaled to fit a 7-point Likert-scale (0 to 6) in order to have comparable scores amongst the different scales and to compute effect sizes. The following equation was used to calculate the Cohen’s d statistics effect size coefficient:
where Spooled equals
where n represents the number of observations (i.e. 4 observations for Motivation/Apathy, and 2 observations for QoL/functioning) and SD the within-person standard deviation. This coefficient thus takes into account the number of observations and the standard deviation within each phase (pre and post, or pre and follow-up) and within each participant.
To further investigate the effects of the intervention, we examined whether Switch had an effect on the ESM variables during the intervention and at follow-up. In a first step, we calculated the means and standard deviations per phase. Additionally, we performed unequal variances t-tests and computed the Cohen’s d statistics (with pooled standard deviation as the denominator) to estimate effect sizes. In a second step, we fitted eight separate autoregressive lag 1 (AR(1)) models using the lm function in R (version 3.6.1). Dependent variables were the outcomes of interest, whereas independent variables were the lag of the dependent variables and the intervention phase, thus representing the autoregressive parameter and the mean intervention effect for each intervention phase respectively. Given that the AR(1) model assumes identical distribution of the errors throughout time, we lagged the independent variable within day and phase, resulting in a missing value for the lag at the first prompt of the day, as well as at the first prompt of a new phase. After each model, we performed a test for homoscedasticity after removing outliers, using the outlierTest function of the car package in R. In case residuals appeared to be heteroscedastic, the initial model (with outliers) was refitted using robust standard errors, by defining the robust variance-covariance matrix as argument using the function vcov of the stats package in R. For the analyses in both steps (t-tests, Cohen’s d, and AR(1) models), baseline scores were taken as the reference group. Finally, we investigated – in descriptive pie charts – how nominal ESM variables evolved throughout the intervention. All analyses were performed for each participant individually.
Aim 2: dynamics between motivation and related processes
To explore the dynamics of the processes associated with motivation, we used individual network representations based on vector autoregressive (VAR) modelling. We computed network models for each phase separately (baseline, intervention and follow-up). We used the graphicalVAR package (version 0.2.2 [37]) to estimate the temporal and contemporaneous models and to obtain a visual representation using the qgraph package (version 1.6.4 [38]). The nodes in the networks represent the variables, whereas the edges (i.e., the lines) represent the associations between the variables. In the temporal network, the edges are directed and indicate which variable predicts other variables in the next timepoint. In the contemporaneous network, the edges represent partial correlations between the variables, after controlling for all the other variables in the same timepoint and also in the previous timepoint. All associations reported and represented on the models are significant (p<.05).
We also calculated plots representing rolling means (or “moving averages”; see Additional file 1, Figure S1). The course of the different processes – during baseline, intervention and at follow-up – was plotted using the rollapply function from the emaph package (version 1.0.0 [39]). This provides rolling means, i.e., the means of each variable as it progresses over time. A rolling mean smooths the time-series, thus making it easier to detect any evolution (e.g., trend and periodicity) and to reveal any associations between variables.