The psychological therapy preferences of patients who hear voices

ABSTRACT Background Voice-hearing is a common, phenomenologically diverse, experience across different mental health diagnoses. Patient preferences for psychological therapies are helpful in informing treatment commissioning and provision, especially in the context of complex and variable experiences like voice-hearing. There is, however, very limited evidence as to the psychological therapy preferences of transdiagnostic voice-hearers. Methods Three-hundred and thirty-five voice hearers were recruited from secondary care NHS mental health services across England, between 2020 and 2022. Participants completed a questionnaire battery, involving a psychological therapy preference survey. Participants ranked their preferences across categories of practical, technical and relational therapy elements. Therapy preferences were examined using non-parametric ANOVAs and the significance of pairwise comparisons between different therapy elements. Results There were significant differences in all categories of preference elements. Clear hierarchies of preference were observed in therapy location, timing, delivery, and therapy approach. Preferences were evident, albeit with less clear vertical hierarchies, for number of sessions, mode, therapist qualities, and therapy focus, tasks and outcomes. Discussion Overall, participants expressed a preference for individual, face-to-face intervention of at least nine sessions, with a highly experienced therapist and a core focus on enhancing coping strategies for voice-hearing experiences.


Introduction
Voice-hearing, also known as auditory verbal hallucinations without appropriate external stimuli, is common for both people with mental health problems and people with no need for psychiatric care (De Leede-Smith & Barkus, 2013). Voice-hearers with a need for care tend to report more, louder, and more negative voices than those without a need for care De Leede-Smith & Barkus, 2013). Explanations for voice-hearing do not reliably distinguish between the two groups; however, people without a need for care more often identify a benevolent spiritual origin (De Leede-Smith & Barkus, 2013). Voice-hearers with a need for care have variable diagnoses, including schizophrenia, bipolar disorder, post-traumatic stress disorder, and emotionally unstable personality disorder De Leede-Smith & Barkus, 2013). Psychotic experiences are associated with risk of long-term social and vocational problems, mood problems and suicide (Kjelby et al., 2015). NICE (National Institute for Health and Care Excellence, 2014) recommend that everyone with psychosis should be offered psychological therapies; specifically Cognitive Behaviour Therapy (CBT) for at least 16 sessions. However, only a quarter to a third of patients are offered CBT (Royal College of Psychiatrists, 2018), seemingly because of limited clinician time (Hazell et al., 2017;Ince et al., 2016). Yet patients' attitudes are important too; at least half of patients with psychosis reportedly decline CBT when offered it (Royal College of Psychiatrists, 2018).
Anti-psychotic medication does not improve voice-hearing for at least a sizeable minority (Aleman & Larøi, 2014), and its use risks adverse side effects like metabolic syndrome (De Hert et al., 2006). Psychological therapies offer an arguably safer and more collaborative form of treatment. Recent psychological therapy innovations for voice-hearing include brief therapistguided self-help CBT for voices, focusing on beliefs about the self and voices ; AVATAR and Relating Therapy, both focusing on relationships with voices and others (Craig et al., 2018;Hayward et al., 2017); and cognitive therapies such as COMMAND, focusing on reducing risk behavior associated with commanding voices (Birchwood et al., 2014, and Mindfulness with Person-Based Cognitive Therapy, which foregrounds the hearer's engagement with, and reactions to, voices (Chadwick et al., 2016). The Hearing Voices Movement (HVM) has grown over the past 20 years as a prominent international approach with associated peer support groups for voice-hearers. HVM challenges traditional biomedical understandings and treats voicehearing as a meaningful human experience (Corstens et al., 2014). Additionally, compassionfocused therapy is gaining attention in its application to voice-hearing; supporting voice-hearers to update their motivational patterns with greater orientation around compassion and safety (Heriot-Maitland et al., 2019). Such treatment and support approaches differ markedly in their foci, techniques and modes of delivery, yet there is limited evidence as to what type of intervention voice-hearers actually want.
The broader psychotherapy literature emphasises that patients want therapists to be attuned to their preferences and goals, and to invite ongoing discussion about this fit (Levitt et al., 2016). The alignment of therapies to patient preferences not only affects therapy uptake, but also satisfaction, engagement, alliance, adherence and outcome (Swift et al., 2019;Windle et al., 2020). NICE recommends exploring patient preferences to inform commissioning and treatment pathway development; especially when considering heterogenous populations and multiple treatment options (Bouvy et al., 2020). Voice-hearing is a complex and variable transdiagnostic phenomenon, with many apparent "subtypes" (Smailes et al., 2015). There is a need to further explore voice-hearing treatment options and patients' own preferences in the context of this clinical heterogeneity .
There are no known studies of patient preferences for voice-hearing intervention specifically, in a psychosis or broader transdiagnostic context. One known study explored barriers and facilitators to implementing a brief guided self-help intervention with transdiagnostic voice-hearers (Hazell et al., 2017); participants emphasised the importance of the therapist being compassionate, empathic, trustworthy, highly-qualified and experienced in working with voice-hearers. A small number of studies have focused on psychosis treatment in general, with two known studies using preference elicitation methods. One study focused on treatment outcome preferences in schizophrenia (Rosenheck et al., 2005). The most preferred outcome target was confusion, followed by energy levels, symptoms, work life, social life, and treatment side effects. Another study focused on treatment type preferences of Early Intervention in Psychosis patients (Welsh & Tiffin, 2014), finding that the provision of psychoeducational material was preferred, followed by CBT, behavioural activation, omega-3 fatty acid supplements and, finally, art therapy. Other studies in psychosis suggest that patients most value treatments that target cognition, mood, physical health and social and occupational functioning (Bridges et al., 2013;Byrne et al., 2010), but with individual variability. Another study suggested half of patients prioritise clinical improvement and half functional improvement (Bridges et al., 2018). There is a need to explore whether such marked individual variability in treatment preferences characterises the opinions of voice-hearers, with and without psychosis diagnoses, as pertaining to voice-specific intervention.

The current study
Widescale surveying of patients' preferences is needed to help refine the focus of research, commissioning and clinical activity, ensuring these all reflect what is important to voice-hearers themselves. We define patient preferences as the relative desirability of different elements and outcomes of healthcare, identified using preference elicitation techniques (Klose et al., 2016). The current study aimed to elicit the psychological therapy preferences of patients who experience voice-hearing.

Design
This study presents quantitative results from PREFER (Patient preferences regarding psychological therapies for voice-hearing experiences), an observational cross-sectional mixed-methods survey study. Lived experience experts were consulted in the study design, and in the development of the patient preferences measure and study recruitment materials.

Participants and procedure
Patients were recruited from inpatient and outpatient youth and adult mental health services across 25 NHS Trusts in England. Participant advertising and informational materials informed participants that PREFER was a large national survey about therapy preferences recruiting people who had heard voices for at least six months. Inclusion criteria required that participants were aged 16 years or over, accessing secondary mental health services with a designated lead practitioner, had sufficient English proficiency for consent and participation, and had heard voices for at least six months, currently or in the past. No exclusions were made with respect to participant diagnosis (presence/ absence or nature), medication use (presence/absence or type), or historical/current exposure to psychological therapy.
Patients completed a self-report questionnaire battery. Quantitative data on preferences, and demographic and clinical characteristics are reported here. The battery was completed with the support of a researcher, via an online questionnaire administered using Qualtrics or on paper, as preferred. The order in which measures were presented was randomised in the online questionnaire. Upon the imposition of COVID-19-related social restrictions, an amendment was made to allow the study to be conducted via telephone and/or video-conferencing. Participants gave consent in writing, or, after the COVID-19 amendment, electronically using a digital consent form.

Preference survey
A preference survey was created for this study. Questionnaire items were derived by mapping items from: psychosis patient preference studies (Bridges et al., 2018(Bridges et al., , 2013Byrne et al., 2010;Rosenheck et al., 2005;Welsh & Tiffin, 2014); psychosis patient experience and opinion studies Greenwood et al., 2010;Holding et al., 2016;Kilbride et al., 2013;Miles et al., 2007); and techniques, processes, outcomes, barriers and facilitators relevant to voicehearing interventions (Hazell, Greenwood et al., 2018;McCarthy-Jones et al., 2014). Items were collated into practical, relational and technical domains and grouped into sets of approximately five items for ranking from most to least preferred (see , Table 1). This forced choice method was selected to ensure a prioritisation of elements relative to each other. Items were refined in consultation with the Sussex Partnership NHS Foundation Trust Psychosis Theme Group; a lived experience expert group who consult on research design and delivery. The group was presented with a summary of the aims and draft design of the study (i.e. a cross-sectional self-report online survey), draft participant information sheet, and draft questionnaire battery, and were invited to propose changes. No study design amendments were suggested. Changes were made to the preference survey, including adding "anxiety" to the preference survey as an outcome target and adding a final free-text qualitative question to capture anything else participants wanted to share about their therapy preferences.

Demographic and clinical characteristics
Participants self-reported their age, ethnicity, gender, employment, and highest educational qualification achieved. Voice-hearing severity was captured using the Hamilton Program for Schizophrenia Voices Questionnaire (HPSVQ (Van Lieshout & Goldberg, 2007)) -a 9-item selfreport measure of voice-hearing symptoms. The HPSVQ has been validated with UK transdiagnostic voice-hearers (Berry et al., 2021). Two supplementary questions were added to assess the presence and co-occurrence of multi-modal hallucinations.

Analysis
Analyses were conducted using SPSS, version 26.0 (IBM Corp, 2019). Descriptive characteristics were calculated using means, standard deviations, frequencies, percentages, minima and maxima. Missing data were examined using t-tests and chi-square tests. Friedman (non-parametric ANOVA) tests were used to evaluate the nature and significance of differences in preferenceranked elements (Arch et al., 2018). Kendall's W was computed, with 0.26 approximating a large effect (Walker, 2003). All pairwise comparisons were produced. All analyses were Bonferronicorrected for multiple testing.

Participant characteristics
In total, 342 participants consented between April 2020 and January 2022 (see, Figure 1); of whom 335 proceeded to the questionnaire. Due to a national embargo on (non-COVID-19) health research recruitment announced in March 2020, most (n = 253) participants were recruited from 2021 onwards. Forty-four participants (13.3%) reported hearing no voices in the past week. Voice-hearing severity scores (Table 1) indicated that the sample experienced moderate to severe voice-hearing on average, with 41% scoring as severe (Kim et al., 2010). Most participants had experienced voice-hearing for over 10 years (Table 1). Further participant characteristics are provided in a supplementary file. Three participants provided only demographic data. There were missing data in at least one preference category for 69 participants (20.6%). Comparison tests indicated this missingness was not significantly associated with any demographic or clinical characteristic (as listed in Table 1 and supplementary information).

Psychological therapy preferences
Overall, patients expressed a desire for psychological therapy and a high likelihood of uptake if offered (Table 1). Most participants reported multi-modal hallucinations and wanted psychological therapy for these experiences (Table 1). There was a highly significant difference in preferences for every element category. Kendall's W values suggested large differences in mean preference rankings for timing of therapy sessions, therapy mode and delivery, and therapist qualities (Table 2). Figure 2 presents all pairwise comparisons; indicating whether each element significantly differs in preference ranking compared to all other elements in the same category. Findings suggest a near complete hierarchy of preferences for location (Figure 2(i)), timing (Figure 2(ii)), delivery (Figure 2(v)) and approach (Figure 2(vii)), for most pairwise comparisons reached statistical significance, meaning that most elements are significantly preferred relative to the next preferred element. The preferred therapy location was a clinical setting (Table 2), followed by own home, community and a variable/flexible location (with no significant difference between these two options), and, lastly, an educational setting. Weekday afternoons were the preferred timing of therapy sessions, followed by weekday mornings, then weekday evenings and weekend days (the mean ranks of which did not significantly differ from each other), and lastly, weekend evenings. Face-to-face was the preferred delivery method, followed by a combination of methods and telephone delivery (which did not significantly differ from each other), then video, and finally text-based therapy. A collaborative therapy approach was most preferred, followed by an evidencebased approach and the avoidance of causing additional distress (with no significant difference between these two approaches). A lack of requirement for long periods of concentration followed, with a lack of requirement to think or talk about the past as the least preferred approach.
Individual therapy (Table 2) was significantly preferred to all other modes (Figure 2(iv)). Guided self-help and group therapy were next preferred, and whilst they did not significantly differ to each other, they were significantly preferred to family therapy or (un-guided) self-help (which did not significantly differ from each other). There was an overall preference for a longer therapy duration (Table 2; Figure 2(iii)), with 9 to 12 sessions most preferred. However, there were no significant differences between 9 to 12 sessions, 13 to 16 sessions, or more than 16 sessions. Fewer than 9 sessions was least preferred. There was no significant difference in preference between more than 16 sessions compared to 4 to 8, however, suggesting an overall curvilinear association between preference and therapy duration.
Preferences were expressed for experience-based therapist qualities above interpersonal and demographic characteristics (Table 2; Figure 2(vi)). The three most preferred qualities were being highly experienced in working with voice-hearers, highly-qualified as a mental health professional, and having personal voice-hearing experience. A high level of professional experience was significantly preferred relative to qualification and personal voice-hearing, and the latter two categories did not significantly differ from each other. Interpersonal qualities followed. Being compassionate and non-judgemental did not significantly differ in mean rank to each other, and being compassionate did not significantly differ in ranking from personal voice-hearing experience. Being hopeful and then a therapist chosen for specific demographic characteristics followed, with both qualities significantly less preferred relative to all proceeding characteristics.
There was a preference for therapy to focus on behavioural and emotional, rather than meaningbased, relational, and cognitive aspects of voice-hearing experiences (Table 2; Figure 2(viii)). The most preferred therapy focus was on coping with voices, which was significantly preferred to a focus on voice-related distress, which in turn was significantly preferred to focusing on voice relationship/ s. A relational focus was not significantly preferred to a focus on ideas about voice origin, nor to the least preferred element, a cognitive focus on assumptions (beliefs) about voices. These patterns were echoed for therapy tasks (Table 2; Figure 2(ix)). Developing and practicing coping strategies was most preferred and significantly more so than all other tasks. Following this, understanding voices and their links with thoughts, emotions and life experiences (i.e. formulation), and identifying and evaluating unhelpful assumptions and beliefs about voices (i.e. cognitive work) were next preferred, and did not differ significantly from each other. The fourth preferred task was therapeutic relationship development, which did not significantly differ to psychoeducation, which in turn did not significantly differ from the least preferred task of developing a sense of one's personal values and goals. There was no clear hierarchy of preference for what outcomes therapy should target (Table 2; Figure 2(x)). The most preferred target was the negative impact on voices, although this did not significantly differ in preference rankings from targeting anxiety or stress, and, in turn, these did not significantly differ from the preference for targeting low mood. Following low mood was self-esteem, which did not significantly differ from the following targets of hopefulness, problems with memory or concentration, or social relationships. Participating in activities, was significantly less preferred to all other proceeding candidate targets except for social relationships.

Discussion
This is the first known study of the psychological therapy preferences of transdiagnostic voicehearers. This study identified a vertical preference hierarchy for practical therapy elements, i.e. location, timing, delivery, and for the therapy approach. For more technical and relational elements, i.e. session number, mode, therapist qualities, and therapy focus, tasks and target outcome, there was typically one clear over-arching preference but limited differentiation amongst lesser preferred elements. Ranking patterns within and across categories overall suggest a preference for collaborative, individual, face-to-face intervention, of at least nine sessions, with a highly-qualified professional with extensive voice-hearing experience. Therapy should foreground coping strategies enhancement and the reduction of voice impact, and general and voice-related distress. Patients preferred therapy to attend to behavioural and emotional, rather than relational and ontological, characteristics of voice-hearing. The extent to which patients prefer a cognitive focus is equivocal; a focus on beliefs about voices was the least preferred focus of therapy, yet identifying and evaluating unhelpful beliefs about voices was the third of six preferred therapy tasks.
Patient preferences for collaborative therapy with a task focus on formulation, i.e. understanding voices and their links with the person's thoughts, actions and experiences, align with the consideration of voice-hearing as a meaningful experience within individuals' life narratives. This is an approach foregrounded in personal recovery models as well as in psychological interventions including cognitive behavioural, relating, and compassion-focused therapies for voice-hearing (Lonergan, 2017) and the Hearing Voices Movement (HVM; Corstens et al., 2014). A previous study evidenced that voicehearers want therapists to be compassionate, empathic, trustworthy, highly-qualified and experienced in working with voice-hearers (Hazell et al., 2017). The present study advances this to suggest that therapist experience and qualification matter most. Such preferences need to be considered within the context of a national shift towards psychosis treatment provision by non-expert professionals (Thomas, 2015). Current findings emphasise that any such "non-expert" should be very experienced in working with and/or have lived experience of voice-hearing, and possess professional qualification/s and specific skills in collaboration and formulation. Professional qualification is not mutually exclusive to lived experience, but is not necessarily typically co-present, and, depending on the approach, may be considered non-complementary. Consideration is needed, therefore, regarding how lived experience experts can work within or alongside packages of care in a way that aligns with patient preferences yet preserves the integrity of the lived experience approach . This might include selecting and/or supporting lived experience experts with professional training that aligns well with their practice, perhaps for example, from the HVM (Corstens et al., 2014). Alternatively, patients may value accessing therapies in which multiple therapists, reflecting among them both professional qualifications and lived experience, play a role. Otherwise, patients could be signposted to access peerrun organisations separately alongside their formal health service use (Kay et al., 2017).
A patient preference for face-to-face intervention was clear. This preference may have been elevated due to inherent isolation in the pandemic context, yet nonetheless should shape future practice, i.e. encouraging the (safe) use of outdoor in-person interactions to supplement any remote care delivery. It is notable too that short durations of therapy were not preferred, although longer durations did not appeal more than a moderate nine to 12 sessions. A previous metaanalysis suggested that fewer than 16 sessions of CBTp is not inferior, with nine as the average number of therapy sessions provided (Hazell et al., 2016). Offering nine sessions would thus appear a good compromise between patient preference, and clinical and cost-effectiveness. Nonetheless, flexibility and variability did characterise preferences for location, mode of delivery, focus, tasks and content; therefore, professionals' ability to tailor and personalise sessions is paramount.
Current participants did not prefer therapy targets explicitly focused on social relationships and activity participation. This is surprising considering the central emphasis on social connection and participation in patient-defined recovery (Andresen et al., 2010) and prior research emphasising patient preferences for social recovery-oriented outcomes (Bridges et al., 2018). These prior preference studies involved people with psychosis and elicited general treatment preferences. Present participants were perhaps inadvertently cued to favour a symptom-focus due to the framing of the study as concerning preferences for voice-hearing intervention. COVID-19-related social restrictions may additionally have resulted in patients being less focused on social recovery and/or more preoccupied with their voices. Moreover, the present sample largely experienced severe and enduring voice-hearing and perhaps thus prioritised symptom improvement and/or perceived coping strategies enhancement as a necessary foundation before focusing on social recovery. Eliciting patient preferences regarding treatment foci and target outcomes in the context of sequential or modular packages would allow further exploration of this possibility. Further research is additionally needed to ascertain how and why patients arrive at their treatment preferences, and how such preferences associate with clinical, demographic, psychological and social factors. This evidence has important implications for the appropriate resourcing and targeting of interventions across clinical settings and stages of recovery.
Research suggests that coping strategy enhancement (CSE) interventions can be effectively delivered to transdiagnostic voice-hearers in routine clinical practice (Hayward, Edgecumbe et al., 2018), including for people with multi-modal hallucinations (Badcock et al., 2021), and with equivalent outcomes in the context of emotionally unstable personality disorder versus psychosis diagnoses (Morrice et al., 2021). Thus, scalable and effective interventions that align with voice-hearers' preferences identified here do exist. Symptom-specific interventions have been identified as resulting in greater effect sizes than when targeting a number of symptoms using broader intervention protocols (Hazell, Greenwood et al., 2018). Yet current findings call into question the extent to which patients value symptom-specific interventions. In addition to wanting interventions that focus on multi-modal hallucinations, the current sample positioned broader experiences of low mood, anxiety and stress as priority treatment targets. Voice-hearing-specific interventions such as CSE can result in broader changes to mood, subjective recovery, and other outcomes (Hayward, Edgecumbe et al., 2018). Therefore, where symptom-specific interventions are the offer, patients should be informed that these interventions do facilitate change in these broader outcomes -and studies of these interventions should aim to capture such wider changes.

Strengths and limitations
The key strength of the present study is the elicitation of ranked preferences across categories of psychological therapy elements. This builds on previous work by identifying not only what matters to patients, but which specific elements matter most. The inclusion of a transdiagnostic sample is another major strength, for voice-hearing is an experience that commonly occurs across diagnoses . Further work should consider how preferences may vary by diagnosis.
There are nonetheless some important limitations of the preference elicitation method. The moderate level of missing data may suggest that ranking preferred elements is challenging to do. The lack of a clear preference hierarchy for categories with more elements may specifically suggest people struggle to discriminate amongst more than five. Equally the invitation to rank different elements implies some independence, which does not necessarily reflect reality. This issue may explain, for example, the low ranking of establishing personal values and goals compared to developing coping strategies. Presumably coping strategies are used at least partly to enable people to pursue valued activities and relationships, yet inviting element ranking perhaps implies they are mutually exclusive. Asking participants to select preferred elements, or to rate each element individually in importance, may better reflect their non-exclusivity; although such methods risk all elements being selected or equivalently rated with no discrimination between them.
More general limitations are of note. This study comprised a self-selecting sample which, although regionally diverse, is predominantly White. This unfortunately reflects the same bias inherent in psychological therapy provision, for White patients are more likely to access psychological therapies for psychotic experiences than Black and Minority Ethnic (BME) patients (Mercer et al., 2018;Morris et al., 2020). Moreover, although there was some variance, most participants expressed a desire for psychological therapy and a likelihood of uptake if offered. If at least half of patients with psychosis decline CBT (The Royal College of Psychiatrists, 2018), then people who feel more ambivalent or disinterested are likely under-represented here. The preferences of this unrepresented group may differ from those with greater desire for psychological therapies, and thus present findings may not help with engaging the former. There might be some important presently unexplored interaction too between ethnicity and uptake. It could be that BME patients are more likely to hold spiritual or otherwise non-medical explanations for their voice-hearing; instead of, or alongside, viewing these experiences as symptomatic (Cook, 2019). If psychological therapies are not seen to be able to take into account the spiritual and religious frameworks within which people experience their voice-hearing (although this can be done (Cook, 2019)), then therapy uptake may be undermined. Future research should aim to explore how preferences differ across cultures and different geographical and health system settings. We did not capture additional patient characteristics such as neurocognition, literary and numeracy, which affect medical intervention preferences more generally (Russo et al., 2019). Finally, as noted, the COVID-19 pandemic and enforced social restrictions may have inflated the preference for face-to-face therapy.

Conclusions
This is the first known study to explore the psychological therapy preferences for a transdiagnostic sample of people who hear voices. Participants expressed clear preferences for practical elements of therapy, with less clear preference hierarchies for more technical and relational elements. Overall, patients want evidence-based, individual, face-to-face intervention of at least nine sessions, with a highly experienced and qualified professional, that focuses on the development of coping strategies, and the reduction of voice impact and general distress. Intervention should be predominantly focused on the behavioural and emotional aspects of voice-hearing, with attention to the more relational and ontological characteristics if expressed as a specific preference of individual patients. Therapists should work collaboratively, be compassionate, and should have skills in formulation. Further research is needed to explore how and why patients arrive at their preferences, and how such preferences may relate to patients' sociodemographic characteristics, their voicehearing experiences, and broader psychological and social factors.