Ultra-brief acceptance & commitment therapy for inpatients with psychosis – a single-case experimental design investigating processes of change

ABSTRACT People with primary psychosis are among the most seen in inpatient psychiatry. Treatment guidelines recommend both pharmacological and psychological treatments. However, psychological treatments are not routinely offered in many settings. There is also a lack of research on psychological treatments for this vulnerable population in the inpatient setting. The first aim of the current study was to examine treatment effects of a brief form of Acceptance and Commitment Therapy on outcomes valued by the treatment recipients. The second aim was to explore hypothetical processes of change in relation to outcomes over time. Three people with primary psychosis were treated for two to four sessions. A replicated single-case experimental design with multiple baselines across subjects (Clinical Trials registration number ID NCT04704973) was employed to examine treatment effects. The Personal Questionnaire (PQ) was used as primary outcome, symptom believability and preoccupation as proposed processes of change. Data were analyzed using visual inspection, calculation of Tau-U values, and cross-lagged correlation. All participants improved significantly on PQ and the symptom preoccupation measure. Two improved significantly on the symptom believability measure. Cross-lagged correlation analyses showed no clear mediation. Change in proposed processes of change and primary outcome predominantly happened concomitantly, although patterns of results reflected individual differences.

gold standard (American Psychological Association, n.d.-b.) among psychological treatments.Recent meta-analyses indicate small-to-medium effects of CBTp for psychotic symptoms, with increasing effectiveness across time for delusions (Sitko et al., 2020;Turner et al., 2020), but evidence about quality of life or functioning is less conclusive (Laws et al., 2018).
As a response to the limitations of disorder-specific treatments, CBT now includes a focus on transdiagnostic processes of change, particularly for enhancing functioning and quality of life (Ong et al., 2022;Schaeuffele et al., 2021).Acceptance and Commitment Therapy (ACT; Hayes et al., 2013) is one example of a process-based approach to behavior change (Hofmann & Hayes, 2019).Its set of processes is called psychological flexibility, defined as the ability to act with openness, connect with the present moment, and change or persist in behavior, based on what one holds as important and how one wants to pursue that.Six processes constitute psychological flexibility: acceptance, cognitive defusion, present-focused attention, self-as-context, values, and committed action (Hayes et al., 2006).
ACT and the psychological flexibility model have been formally applied to psychosis (ACTp; Bach & Hayes, 2002;Gaudiano & Herbert, 2006).Meta-analyses of ACTp show moderate to strong effects for reducing psychotic symptoms, rehospitalizations (Cramer et al., 2016;Tonarelli et al., 2016), and improving affective symptoms (Yip et al., 2022).Individual trials show that ACTp appears acceptable and feasible in inpatient care (Tyrberg et al., 2017b), and can also improve quality of life, and, again, prevent rehospitalizations (Bach & Hayes, 2002;Gaudiano & Herbert, 2006;Tyrberg et al., 2017a).Recent reviews (Louise et al., 2018;Wakefield et al., 2018) report promising results for ACTp, but also call for research that is higher quality and addresses new questions.ACTp is deemed as having a modest level of research support by the American Psychological Association (n.d.-a.).
The psychological flexibility model is supported by experimental evidence (e.g.Levin et al., 2012), but more studies are needed (Morris et al., 2023).Symptom believabilityconsidered as an aspect of cognitive defusion (Zettle et al., 2011)-has been found to mediate distress with psychotic symptoms (Gaudiano et al., 2010).Preoccupation with psychotic symptoms has been found to predict depressive symptoms and need for care in psychosis (Sisti et al., 2012;Thomas et al., 2014).Theoretically, psychological inflexibility certainly includes preoccupation with internal experiences, and this can exacerbate the impact of psychotic symptoms (Thomas, 2015).
Previous studies of ACTp rely on aggregated group data and derive summary mean group effects.It is argued that to improve treatments, we should adopt a greater focus on individual people, so called idiographic approaches, in looking at both outcome and processes of change (Fisher et al., 2018;Kazdin, 2016).
ACT appears to produce benefits on average in inpatient contexts (Bach & Hayes, 2002;Gaudiano & Herbert, 2006;Tyrberg et al., 2017a), but we know less about how particular individual people respond to ACTp and which processes of therapeutic change underlie their results.The aim of the present study was to extend previous work by exploring the effectiveness of a brief ACTp intervention in an inpatient setting, using an individualized approach to both treatment delivery and research, with the following research questions: (a) Does ACTp produce changes in the person's specific psychological difficulties?(b) Does ACTp produce changes in the dimensions believability and preoccupation with psychotic symptoms?(c) Do believability and preoccupation mediate the change in the primary outcome?

Design
The study employed a Single Case Experimental Design (SCED).We used a replicated AB-design (where A is baseline before intervention, and B is intervention phase) with nonconcurrent multiple baselines across subjects and employed additional pre-post measures.The lengths of baselines varied to the extent that it was feasible, with a minimum length of five points, following the Single-Case Reporting Guidelines in Behavioral Interventions 2016 (SCRIBE; Tate et al., 2017) and a minimum difference of two points as recommended by Morley (2018).The intervention phase began with the first ACTp session and concluded after the last.Measures of outcomes and possible mediators were taken four times daily during all phases.Pre-post measures were taken before the first ACTp session, and after the last.Blinding and randomization of baselines were not possible because of limitations imposed by the clinical setting and the nature of the intervention.
The study protocol was pre-registered on https://ClinicalTrials.gov (ID NCT04704973).The record describes a larger study that was subsequently divided into two studies, the present is the second one.The mediation analysis for the present study was not pre-registered.Ethical approval was received by the Swedish Ethical Review Authority (2020-03251).

Participants and setting
Recruitment for the study took place between June -December 2022.Inclusion criteria were: (a) admitted to one of the inpatient wards of the psychiatric clinic where the study was conducted, (b) age 18 or over, (c) diagnosed with psychotic disorder according to ICD-10, code range F20-F29 (World Health Organization, 2016).Exclusion criteria were (a) moderate to severe cognitive deficits (routinely assessed by the ward team, including attending psychiatrist, psychiatric nurse and ward psychologist), (b) ongoing participation in outpatient psychological treatment, (c) significant ongoing planned changes in the pharmacological treatment, (d) ongoing somatic disease that causes symptoms of psychosis, (e) substance-induced psychosis, (f) inability to give informed consent, (g) alcohol and drug use disorder, (h) unstable baseline.
The setting was two psychiatric wards, with 12-bed capacity each, situated in a Swedish urban hospital.Participants received treatment as usual (TAU) throughout the study, consisting of pharmacological treatment, observation of symptoms and support for nutrition and sleep.Psychological treatment was not routinely offered on the wards.

Procedure
Participant A was included in the study four days after admission, participant B one day after, and participant C two weeks after admission.All had previous psychiatric inpatient care.None had previous experience of ACTp.Following written informed consent, an assessment interview and Personal Questionnaire (PQ) formulation was administered by the first and third authors (a master's level psychology student and a doctoral level clinical psychologist).Measures were taken in the morning, around lunch, in the afternoon and in the evening during baseline and treatment phases.Two participants received two ACTp sessions and one received four.Sessions were 45 minutes.The assessments continued four to five data points after the treatment phase (Morley, 2018).

Primary outcome, the personal questionnaire (PQ)
The PQ (Shapiro, 1961) is an idiographic measure with statements based on patients' main psychological problems.PQ is an evidence-based, norm-referenced measurement of client psychological distress (Elliott et al., 2016), with a reliability of .82(Morley, 2018).A description of the participants' PQs can be found in Appendix A, with three main problem areas for each patient and statements reflecting "illness", "improvement" or "recovery" in each area.The PQ is sensitive to inconsistent responses, the scores are 1-4 for each problem area-higher scores representing worse health-and 3-12 considering the three main problem areas together (see Morley [2018] or Shapiro [1961] for a complete description of PQ construction, administrations and scoring).

Hypothetical mediators
Symptom believability.Following previous research (Bach & Hayes, 2002;Gaudiano et al., 2010;Peters et al., 2004;Zettle & Hayes, 1986) symptom believability for delusions was measured with the item "how true do you believe [specific delusion] to be?" and a numerical rating scale from 1= "don't believe it's true" to 5= "believe it's absolutely true".This question was taken from the 21-item Peters et al. Delusions Inventory (PDI;Peters et al., 2004).It was not validated as such, but adapted to this study, since it was based on the participants' delusions.In the case of hearing voices, the question was modified to "how much do you buy what the voices say?" with the same numerical rating scale.
Symptom preoccupation.Symptom preoccupation was measured with the question "how often have you been thinking about [specific delusion]?"and a numerical rating scale from 1= "hardly ever think about it" to 5= "think about it all the time", also taken from the PDI.We added a question to assess preoccupation with voices: "how often have you been thinking about [specific voice]?" using the same corresponding numerical scale.

Pre-post measures
The following measures were employed to provide descriptions of functioning before and after treatment.The Bull's-Eye Values Survey (BEVS; Lundgren et al., 2012) measures values-based action, associated with psychological flexibility.Total scores range between 4 and 28, higher scores indicating more values-based action.BEVS has shown satisfactory validity and good test-retest reliability (r =.85).
The Psychotic Symptom Rating Scales (PSYRATS; Haddock et al., 1999) is a 17-item multidimensional measure of delusions and auditory hallucinations with acceptable inter-rater reliability and validity.Only the auditory hallucinations subscale (AHS), consisting of 11 items, was used in this study.As each item is rated on a scale from 0 to 4 total scores range between 0 and 44, higher scores indicating more symptoms.All the participants' main complaints at baseline included hearing voices.

Treatment
The treatment is one proposed by Tyrberg and Gaudiano (2023), based on previous empirical evidence (Bach & Hayes, 2002;Gaudiano & Herbert, 2006;Tyrberg & Klintwall, 2022;Tyrberg et al., 2017a) and clinical experience.In previous publications (Tyrberg & Klintwall, 2022) it has been termed transdiagnostic ultra-brief behavior therapy (TUBB) but is here applied to psychoses specifically.The theoretical basis is a version of the ACT model known as open, aware, active (Hayes et al., 2011), seen as a condensation of the original six processes into three pillars.
The treatment consists of three components.The first is a clinical functional analysis, aiming to clarify the long-and short-term consequences and overall workability of coping strategies the patient has used thus far.Four questions, seen as a template for focused interviewing, guide this analysis (Strosahl et al., 2012): "What are you seeking?", "What have you tried?", "How has it worked?","What has it cost you?".
The second component is an experiential exercise based on the functional analysis: the "lifeline" (Dahl et al., 2009).The lifeline aims to capture how avoidance pulls the person away from what is meaningful in life for them, while active responding also means encountering and carrying painful inner experiences.
The third component is meant to function as a maintenance tool for when the patient is discharged from the ward, and to consolidate what has been learned during therapy.It consists of an experiential exercise called the pause (R. Harris, personal communication, September 19, 2015).The pause offers a repetition of the central steps of the previous work and is distributed on a printed pocket-sized card (Appendix B).The exercise consists of the following steps: pause and control breathing, apply the observer's perspective, experience feelings and thoughts as they are, and choose a direction and act.The patient is encouraged to first practice the exercise while on the ward, and then to keep using it when discharged.
The intervention was carried out by the third author.

Assessment of treatment effect
To determine if changes in primary outcome (PQ) and proposed mediators (believability and preoccupation) occurred due to the intervention, visual inspection and complementary statistical analysis were used (Manolov et al., 2014).Visual inspection follows an examination of the stability of data within a phase, the trend and immediacy of effects (Kratochwill et al., 2013).Tau-U was calculated to examine the degree of overlap between phases and possible baseline trends (Parker et al., 2011;Vannest et al., 2016).Tau is a nonparametric rank correlation coefficient whose value (between +1 and −1) characterizes the degree of agreement between two ordinal variables.Tau-U is a quantitative approach for analyzing SCEDs.It combines nonoverlap between phases with intervention phase trend and can correct for a baseline trend (Parker et al., 2011).Identifying trend involves comparing the baseline phase with itself.Tau-U provides an index of improvement.Regarding the interpretation of the effect size of the intervention, a 0.20 improvement may be considered a small change, 0.20 to 0.60 a moderate change, 0.60 to 0.80 a large change, and above 0.80 a large to very large change (Vannest & Ninci, 2015)

Analyses of processes of change
To determine potential processes of change in treatment response we used cross-lagged correlation analysis, to look at both temporal association between the potential mediators and primary outcome, and strength of association, after adjusting for autocorrelation (Borckardt et al., 2008).To perform the analysis Simulation Modelling Analysis was used (http:// clinicalresearcher.organd https://posit.co/download/rstudio-desktop/).Correlations between the PQ and hypothetical change processes were estimated and compared at lags − 7 to + 7, after testing and not finding significant correlations at superior lags than 6 for any participant.
Negative lag denoted a change in the proposed process before PQ.Positive lag denoted a change in the process after the primary outcome changes.A zero-lag indicated that the primary outcome changes contemporaneously with the proposed process.The number of simulations was set at 10 000.The significance level was set at p < 0.05.

Participant characteristics at baseline
Six participants were included in the study, all admitted voluntarily.Three began but dropped out of the study for different reasons.One of them found the repetitive measures distressing but wished to continue with the therapy sessions.The second was discharged before treatment could commence due to the absence of beds on the ward.The third showed an unstable highly variable baseline.Three participants remained in the study.
See Table 1 for a description of each participant's characteristics.

Housing with staff
The dose of the pharmacological treatment was reduced during the baseline.
All participants completed the treatment.Participant A was included in the study in September 2022, participants B and C were included in October and November 2022, respectively.Missing assessment points are included in the cross correlation analyses.

Effect of the intervention on primary outcomes
For all participants, visual analysis indicated reduction in the distress caused by the main problem areas.Participant B showed a rising trend, reflecting deterioration, at the end of the treatment phase, from assessment number 33 onwards (Figure 1).TAU did not change during the assessment for any of the participants.Patients A and C did not experience further stressful life events during the time of the study, but participant B coped with the loss of a loved one after assessment 27 onwards.
The complementary Tau-U analyses indicated that, using an analysis that is robust in circumstances of baseline trends, most of the data from baselines and treatment phases were non-overlapping and phase-dependent (between 89% and 99.3% non-overlap).The change was statistically significant in all cases and on average for the group (see Table 2 for details), with large effect sizes (Parker & Vannest, 2009).Hence, both analyses suggest that the intervention was effective (i.e. less distress and greater improvement in the outcomes valued by the person) in terms of behavioral change for all the participants.
Complementary measures BEVS and PSYRATS/AHS support the presence of a treatment effect (see Table 3).

Effect of the intervention on potential processes of change
Visual inspection (Figure 2) and statistical analysis (Table 4) indicated a significant treatment effect for preoccupation in all participants.Participant B showed again a rising trend at the end of the treatment phase.The same analysis (Figure 3 and Table 5) showed that for believability there was an improvement in participants A and C, but not in participant B. Again, we note that the test being used here can accommodate baseline trends.
The results of the cross-lagged correlation analysis between the selected hypothetical process of change and the primary outcome PQ are shown in Table 6.For participants A and B, the changes in preoccupation and believability occurred at the same time as changes in the primary outcome (at lag zero), both correlations are significant.For participant B, there is also a weaker but significant correlation between preoccupation and the primary outcome at lag +1.For participant C, significant correlations between both processes of change and the primary outcome happen at lag −6 and lag 0, with the strongest correlation at lag 0. This suggests that the changes in preoccupation and believability occurred concomitantly or preceded the changes in the primary outcome.In Figure 4 it is shown how the process of change unfolds for each participant, with the lagged correlations for the three cases and two processes.

Discussion
The first aim of this study was to evaluate if the treatment delivered is effective for people with primary psychosis, in a psychiatric inpatient context, in terms of outcomes relevant      to the person.The results show improvements in the primary outcome suggesting a potential treatment effect.This was supported by visual inspection, and by complementary statistical analysis.Reductions in symptom preoccupation for all participants also occurred, and reductions in symptom believability occurred for two out of three participants.It can be argued that these changes happened due to TAU.However, all participants presented high degree of psychological distress (PQ), symptom preoccupation and symptom believability during baseline.These findings are consistent with prepost measures.Reductions in symptom believability only occurred for the complaint of hearing voices-as for participants A and B-but not for delusions-as for participant B.
The second aim was to explore how change unfolds, by exploring how measures of hypothetical processes of change relate to outcome over time.The statistical analyses determined that, for all participants, the proposed mediators (preoccupation and believability) were most strongly associated with outcome at lag zero, meaning that changes in proposed mediators occurred concomitantly with changes in the main complaints (distress due to voices, anxiety, suicidal thoughts, etc.).This may however suggest that one does not causally affect a change in the other.
Participants A and C experienced significant changes after the first session with gradual improvement along treatment and post-treatment phases.Participant B also experienced improvements after the first and second sessions, however the scores worsened after session 3 and 4. As shown in Figures 1 and 2, the PQ scores increased, the same occurred with preoccupation scores.The treatment sought to create behavior change in the participant by targeting relevant processes, in this case believability and preoccupation.Results for participant B could hypothetically mean that targeting defusion might have ensured more solid changes in the primary outcome and other processes.An alternative hypothesis is that an external coincidental event, the loss of a relative, had an impact on the measures of distress and preoccupation.

Strengths and limitations
The instrument to measure the primary outcome has the benefit of being both idiographic and quantitative.It has also been shown to have good psychometric properties and be an appropriate instrument for psychotherapy practice and research (Elliott et al., 2016).Including several problem areas increases the reliability of the instrument (Morley, 2018) and lessens the need to rely on other constructs since problems are expressed in participants' own words.However, constructing the instrument requires time and effort.There is also the risk of statements changing over time.A way to enhance the validity of the instrument was to consult senior colleagues and medical history before deciding which statements were more representative.The two numerical rating items used to measure preoccupation and believability were useful because they were short, as was the PQ, requiring just a few minutes to administer.However, they had the limitations of being self-reports, and had not previously been validated.There is a need to validate measures of interest in this specific population, and to develop and use validated behavioral measures of ACT processes rather than just relying on self-reports.
A further limitation is the risk for expectancy effects, as one of the authors was also the therapist in the study.To minimize this risk, most of the assessments were carried out by the first author.By alternating statements between the top and bottom of each PQ card, the measure itself is also designed to prevent irrelevant response sets.
There are no concrete recommendations about the frequency of administration of the PQ and other items, whether they are adequate instruments to use several times daily.The data obtained in the study shows that they appear sensitive to change.On the positive side, the repetitive measures can prevent the participant from relying on memory, and possible risks of recency or other priming effects, when the measures are taken only once a day or weekly.A potential threat to internal validity is the intimate contact between the delivery of measure and delivery of treatment.
Regarding treatment fidelity, not following a set protocol represents a threat-we cannot unequivocally confirm that all participants received precisely the same treatment, even when the principles of the intervention were the same for every participant.We might argue however, that the treatment was functionally the same even if it differed formally in some ways.
The present study only measured potential effects during patients' stay on the ward.This obviously prevents us from evaluating potential long-term effects of treatment.In future studies, adding long-term follow-up would add important information to these results.
A strength of this study is the multiple-baseline design, the replication across cases, and that the treatment phase was introduced at different times after admission for each participant.A limitation with our design is that all baselines are short in terms of days.The context and ethical concerns made it difficult to create longer baselines.The length of the treatment phase was also variable.More participants with comparable treatment phases (in terms of content, length, and frequency of the sessions) could reduce this limitation.Another limitation was the lack of blindness and nonconcurrent baselines.
SCEDs are particularly vulnerable to alternative rival hypotheses unless they incorporate features needed to immunize against these.We tried to avoid a confounding role of pharmacological treatment in relation to outcome.Participants A and C received the same treatment for two and four years respectively.The exception was participant B, whose pharmacological treatment was changed when being admitted to the ward.However, the participant reported that the reduction in distressing voices did not concur with the change of pharmacological treatment but with the admission to the ward.It is unclear if the new treatment with risperidone contributed to the improvement of the other complaints.TAU also includes other contextual interventions such as routine, consistency and separation from life stressors.Time in the unit could therefore have had an impact on our dependent variables.
This study followed the recommendation to analyze the data with a combination of visual inspection and statistical analyses (Manolov et al., 2014).Tau-U is a robust non-overlap method for this kind of design (Parker et al., 2011).The cross-lagged correlation analysis adjusted for autocorrelation, as a complement of the visual examination of the result, can help in understanding the process of change: how, to what degree, and in what order the different processes are associated in time.The fact that the changes in the processes did not predominantly occur after the changes in the primary outcome suggest that believability and preoccupation remain plausible as hypothetical processes of change.

Conclusion
This study explored the therapeutic potential of including individually delivered ACTp in the ward environment.It demonstrates that a brief intervention focused on enhancing psychological flexibility predicted changes in outcomes valued by the individual patient.The design and complementary analyses in the study allowed us to better understand how factors supporting treatment response and the outcome occurred for each participant.A key lesson from close inspection of the data is that the treatment studied here is substantially different in process and outcome for each person.

Disclosure statement
No potential conflict of interest was reported by the author(s).

Figure 1 .
Figure 1.Graphical display of PQ data across baseline and treatment phases.PQ scores represented on the Y-axis.Arrows pointing out sessions.PQ = Personal Questionnaire.

Figure 2 .
Figure 2. Graphical display of preoccupation data across baseline and treatment phase.Symptom preoccupation represented on the Y-axis.Arrows pointing out sessions.

Figure 3 .
Figure 3. Graphical display of believability of symptoms data across baseline and treatment phase.Symptom believability scores represented on the Y-axis.Arrows pointing out sessions.

Figure 4 .
Figure 4. Graphical display of the lagged correlations for the three cases and two processes.Cross lagged correlations represented on the Y-axis; lags represented on the X-axis.

Table 1 .
Characteristics of participants at baseline.

Table 2 .
Tau-U analysis of treatment effect on PQ.

Table 4 .
Tau-U analysis of treatment effect on process symptom preoccupation.

Table 6 .
Significant cross-correlations at lag zero, lag 1 and lag − 6, indicating the level of association between each proposed process of change and PQ, and respective p-values.

Table 5 .
Tau-U analysis of treatment effect on process symptom believability.