There were four main findings corresponding to (1) efficacy, (2) compliance analyses, (3) psychological suicide mechanisms and (4) acceptability and implementation. These key findings were based on convergent quantitative, qualitative and mixed-methods approaches across all the CARMS workstreams.
Main efficacy finding
The main result from the ITT analysis was that there was no difference between the treatment (suicide-focused CBSPp therapy plus TAU) and control (TAU alone) groups on the primary suicidal ideation severity outcome measure, the ASIQ, at 6 months using the alpha level cut-off of 0.05. This contrasts, at least to some extent, with earlier pilot work with our suicide-focused therapy when trialled in community residents with schizophrenia42 and in male prisoners,53 where the intervention was significantly more effective than TAU. Of course, this may be due to sample differences. Indeed, the primary outcome baseline suicidal ideation mean scores in the current CARMS work were around 20 points higher than in pilot samples.42 Hence, that suicidal ideation became less severe in both groups over 6 months in the current study was, perhaps, unsurprising, given that large regression to the mean effects were probably operating in both treatment and control groups. As this was an actively suicidal sample at recruitment, there was a large potential for suicidal ideation to improve across both trial groups,152 meaning that any therapy effect had to be strong to demonstrate a significant difference compared to TAU. After 12 months, once treatment ended, there were no significant differences between the two RCT groups.
Compliance modelling examining the effects of a ‘dose–response’ and systematic exclusion
Planned compliance modelling used two approaches. The first approach modelled a therapy ‘dose–response’, defined as number of therapy sessions attended, and showed that suicidal ideation severity was reduced by 0.46 of a unit for every additional therapy session attended. The second approach comprised a systematic exclusion analysis, performed by removing from the regression models those participants who had received < 1 therapy session. However, neither of these findings were significant at the 0.05 level.
Determining the effect of CBT ‘dose’ on efficacy in people with non-affective psychosis has been attempted only to a small extent in the literature. For example, one study examined effects of CBT at 5, 15, 25 and 45 sessions and found that by session 15, a stable reduction in psychotic symptoms and associated distress was evident, which appeared to reach a minimum by session 25.153 In another example, analyses methods akin to those used in the CARMS project afforded a more fine-grained approach and showed that each session of CBT therapy reduced a measure of psychotic symptoms and general ‘psychopathology’ by 0.6 points.154 To advance work examining therapy ‘dose’, (1) more advanced statistical modelling analyses are needed to determine the defining characteristics of any subgroup of participants who responded particularly strongly to the intervention and, more importantly, to identify characteristics of participants who responded the least positively, (2) temporal pathways tracking suicidal ideation severity month by month are required, but are, of course, exceptionally resource-demanding to implement and (3) there needs to be a consideration of how therapy ‘dose’ becomes integrated with other therapy process measures, such as formulation complexity and perceptions of the therapeutic alliance, within mechanistic pathways.155
Psychological suicide mechanisms
Based on a robust, albeit continually evolving, evidence base,21,43,55,68,156–160 it was expected that six different types of appraisals would be central in the pathways leading to suicidal thought severity, which were appraisals of emotional difficulties, poor personal problem-solving, poor social support, defeat, entrapment and hopelessness. It was predicted that the suicide-focused therapy intervention (CBSPp) would impact suicidal ideation severity both directly and indirectly via these six types of negative appraisals. In accord with expectations, all six types of appraisals were associated with suicidal ideation severity at month 6, while controlling for baseline scores. However, only one mediated pathway was significant, which was that the suicide-focused therapeutic intervention reduced suicidal thought severity over 6 months not directly, but indirectly, by improving perceptions of poor social support from friends, family and other people more generally (i.e. the mediator). Our qualitative mechanistic findings deconstructed this mediational pathway in more depth in that they underscored that there were three dominant aspects of the interplay between psychotic and suicidal experiences reflecting social interpersonal relationship dynamics. These were (1) feelings and perceptions of not mattering versus mattering, (2) perceptions of connection versus disconnection and (3) the freedom versus the constraints involved in sharing mental health problems.46
Not mattering (sometimes termed ‘anti-mattering’), feeling inconsequential and feeling unimportant have been shown to substantially contribute to mental health problems across the lifespan140,161–166 but have received less attention in the suicide literature.138,139,161 The converse has also been found, which is that perceptions of mattering have been linked with psychological well-being and psychological resilience.140,162,167 Some participants in the CARMS qualitative work described a shift in their thoughts and feelings, wherein their lives were felt to have meaning; they felt they had a ‘role’; and that they did have a sense of mattering not only in the here-and-now but also with respect to the past and looking towards the future. These realisations changed their view of suicide as being a desired option for them. From the perspective of developing interventions, these kinds of descriptions of change and transition represent a potentially realistic goal or value that can be worked towards and actualised. Focusing on ‘what matters’ to a client presents a client-led concrete emphasis.
Feelings and perceptions of being disconnected from other people were described as precursors to suicidal thoughts and behaviours in the qualitative work. This is consistent with a large and growing literature which focuses on perceptions of being a burden not only to other people but also to communities and societies more broadly, in people who have suicidal experiences.22,168,169 Moreover, recent work has linked perceptions of being a burden and not belonging to perceptions of not mattering.170 Our qualitative work showed that fundamental to disconnectedness were perceptions of not being understood, being un-understandable and of other people not seeming to want to understand. These sorts of perceptions could be further amplified if participants felt that their psychotic experiences were being dismissed and/or not recognised, especially if they had found mental health services difficult to access. Our qualitative findings not only expand our understanding of the mechanisms underpinning suicidal experiences but also link in with burgeoning work documenting the damaging, and often hidden, influences of assumptions, judgements and stigma on people with mental health problems.171–173
Acceptability and implementation findings
Participants were deliberately recruited into CARMS to have frequent and/or severe suicidal experiences (e.g. thoughts, urges, images, compulsions, plans and attempts). Both primary and secondary suicide outcome measures attested to the suicide severity of the sample. One concern at the outset was that working with people who were acutely suicidal might be ‘triggering’ and aggravate or worsen suicidal experiences, especially as suicide was assessed intensively using different tools (e.g. questionnaires, a type of TLFB diary, qualitative interviews). Using mixed qualitative and quantitative methods to address this concern, we found that although participants did disclose that talking about suicide could be distressing and a difficult topic, this was balanced by (1) an expectation that sharing experiences would be difficult, (2) being able to gain an increased awareness and self-understanding, (3) valuing the possibility that participation would help other people and (4) a type of catharsis which was long-lasting.115 These positive findings concur with a broader literature mainly examining the effects of participating in suicide research projects which were not RCTs.134,143–151 Taken together, this work counters concerns about ‘suicide talk’ from ethics committees, mental healthcare professionals and suicide research groups.115,174
Given the severity of suicidal experiences of CARMS participants, it was expected that problems relating to therapy adherence and engagement, developing and maintaining the therapeutic alliance, and developing a shared understanding of the foci and processes of therapy (i.e. formulation) might be problematic, especially if participants had felt stigma and/or threatened by involuntary detainment during previous encounters with mental health services.175–178 In contrast to these expectations, evidence for therapeutic adherence and engagement was positive from CARMS. The median number of therapy sessions attended was 16, which aligns with NICE recommendations for CBT for psychosis179 and contributes positively to discussions about the feasibility of suicide-focused therapy delivery in mental health services.
As has been found previously, the therapeutic alliance was higher as rated by participants compared to therapists,180 and did not diminish over time.181 Furthermore, ratings of the alliance were comparable to other samples of people with psychosis recruited to RCTs testing psychological interventions. Therapists used a manual for guidance when delivering our suicide-focused therapy, CBSPp,1 which was monitored in weekly supervision sessions supplemented with peer supervision. Guided by the manual, the mean level of formulation complexity, as rated by the therapists, was found to be at the advanced end of a formulation spectrum, again, testifying to the potential benefits of a suicide-focused therapy182 and the ability of participants with severe mental health problems, specifically non-affective psychosis, to engage with formulation-based therapy processes.
The suicide-focused therapy (CBSPp) was found to be acceptable to participants when views were probed with qualitative interviews.116 All participants had been sampled so that they had recent suicidal experiences and met diagnostic criteria for ICD-10 non-affective psychoses, the most frequent of which was schizophrenia. Positive and negative psychotic symptom scores were equivalent to that reported for psychiatric inpatients2 and community samples.42 In other words, all CARMS participants had severe mental health problems which were no less severe than reported previously and, in the case of suicidal thoughts, were more severe. Yet, despite experiencing severe, persistent and distressing psychotic symptoms, together with severe suicidal ideation and behaviours, participants found the therapy to be acceptable and practicable and were able to engage and persist with one-to-one therapy. These findings echo previous work by our group illustrating the acceptability of suicide-focused therapy in those who were inpatients on hospital psychiatric wards2 and who were incarcerated in male prisons.64 That said, individuals who put themselves forward for qualitative interviews may well be particularly predisposed towards the positives of therapeutic interventions.
Evidence regarding implementation of suicide-focused therapy in NHS mental health service settings was similarly positive and was generated from (1) an examination of SAEs, (2) qualitative analyses117 and (3) exploratory health economics data. Corresponding to work in inpatient psychiatric ward settings,2 no SAEs were considered to be due to the suicide-focused therapy intervention. From an implementation perspective, one concern was that a suicide-focused therapy may ‘trigger’ a greater frequency of suicidal and self-harm behaviours and, relatedly, may have led to more individuals attempting suicide, which would, in turn, lead to a greater demand on mental health resources. We used rigorous procedures to identify and classify SAEs. The majority of SAEs were expected with a similar number of participants across the two trial groups being involved in those constituting suicide attempts, considered life-threatening and being categorised as severe. This seems somewhat reassuring in the context of mental health service implementation of a suicide-focused therapy.
Qualitative CARMS findings examining the facilitators and barriers to implementation converged with our previous work in psychiatric inpatient settings2,75,176,183 in that (1) both mental health professionals and SUs expressed that suicide-focused therapies were urgently needed, (2) mental health professionals relayed that bespoke training for working with people with suicidal experiences was lacking in multiple areas and (3) all participants discussed experiences of available mental health services being overly complex and inhibitory, which may accentuate feelings and perceptions of being irrelevant and not mattering, as underscored by one of our CARMS workstreams.46 This echoes the findings of previous studies which have also highlighted problems with the complexity of available mental health service provision both within and outside of the UK184–186 and called for the active involvement of SUs and carers in the co-design and provision of mental health services.187–189
The health economic aim and analyses were exploratory. Health and social care resource use in the treatment group was estimated to be more expensive than in the control group, albeit not significantly so, and with the observation that the CIs were large, signifying substantial uncertainty in the data. Resource use was based on self-reports, which can often be difficult for participants to remember, especially over a period of months, warranting interpretative caution. Furthermore, we were only able to collect information about health and social care resource use for the 6 months preceding recruitment and 6 months prior to the 12-month follow-up (i.e. when therapy had ceased for the treatment group). Hence, an assumption was made that resource use was comparable in the baseline to 6-month time window, which may not be tenable. Finally, because the health economy analyses were exploratory, they were underpowered.
Limitations
There were five key limitations of the CARMS project which warrant discussion. First, the participants were predominantly White/Caucasian, recruited from the NW of the UK and purposively sampled with respect to suicidal thoughts/acts. This may limit generalisability, especially as NHS Trust operations and service provision can vary depending on geography. Relatedly, a concern with the rigors of psychological intervention RCTs is that the inclusion and exclusion criteria are so strict that they can be ill-matched with characteristics of health and mental health SUs in ‘real-life’.190 The profile of the participants recruited into CARMS negates this concern, at least to some extent, because participant’s experiences reflected the ‘real life’ of mental health services. For example, suicidal thoughts and attempts are highly prevalent in those with mental health problems,16,191 and many people on psychiatric inpatient wards have suicidal experiences.2,192–194 That said, cultural and ethnic perceptions of suicidal and psychotic experiences could not be examined in CARMS, nor could the experiences of harder-to-reach individuals and communities, which should be redressed by future work.195–198 Furthermore, we were not able to include any measures of social and economic deprivation, which is a strong contributory factor to suicide attempts and deaths.199
The second limitation is that the primary suicide outcome questionnaire measured the severity of suicidal thoughts and not suicide attempts or acts. Suicidal thoughts are a strong predictor of concrete suicide plan formation and suicide attempts, including in people with psychosis.200 Furthermore, suicidal thoughts often emanate from immense psychological pain and distress, which, in itself, is an important clinical target.1 The measure used80 was designed to be sensitive to a number of different aspects of suicidal experiences: (1) concrete suicidal thoughts and plans, including writing a suicide note, (2) passive wishes not to be alive, (3) hopelessness, (4) suicide as the only means to solve or escape from problems and (5) the effects of suicide on other people. Nevertheless, aspects of suicidal experiences which are more akin to compulsions or urges,201 involve intrusive images202 or reflect dynamic fluctuations may not have been captured. Furthermore, while many risk factors for suicidal thoughts have been identified, they have poor predictive utility in identifying which individuals will die by suicide.4,203,204 Perhaps one point to note is that the chosen measure of suicidal ideation severity was more sensitive to a treatment effect than the secondary suicide outcome measures. Determining which factors were responsible for this sensitivity is important for future work from pragmatic, clinical and mechanistic perspectives.
A related and third limitation follows, which is that none of the primary or secondary suicide outcome measures were developed by EBEs. The pros and cons of measures, which have been developed by SUs in the mental health arena, have been documented.205–209 Currently, to the authors’ knowledge, such a measure has not been published in the suicide literature.
Fourth, although we had a RA who was specifically assigned to qualitative work, including the analyses, and the CARMers also contributed to qualitative analyses in particular, we were not able to employ a RA who was also a SU. Clearly, this is an area that warrants improvement in future work.
Fifth, only two follow-up time points were used of 6 and 12 months. Ideally, tapping into suicidal experiences more frequently using diary methodologies (e.g. experience sampling210,211) would greatly facilitate a finer-grained understanding of the mechanisms underlying suicidal experiences, but these methodologies are often impractical in the context of a RCT.
Strengths
There were six main strengths of the CARMS project. Despite calls for psychological interventions to be developed that are specific for suicidal experiences,1,55 this has been realised only to a small extent,59 but see.212 The CARMS project was unique in that in order for participants to be included, not only did they have to have/have had suicidal experiences in the past 3 months prior to recruitment, an ICD-10 diagnosis of non-affective psychosis was also required. Work which develops psychological interventions for suicidal experiences has sometimes purposely excluded individuals with active psychosis because therapy can be seen as incompatible with the types of emotional and cognitive manifestations which can accompany psychosis.136 That CARMS proactively sought people with psychotic experiences, such as hallucinations, delusions and paranoid thoughts, beliefs and feelings, who were also recently suicidal was the first clear strength of the current study. Taken together, the findings from CARMS illustrate that such individuals should not be excluded and that suicidal and psychotic experiences are not a contraindication for therapy, but quite the contrary.
The second main strength was PPIE. The CARMers, a self-named SURG, contributed to a diverse array of activities in the phases leading to funding being granted and thereafter throughout the ‘life-cycle’ of CARMS. An important contribution of the CARMers has been presenting their experiences of being an integral part of CARMS in two academic conferences. Calls for genuine collaborative input to mental health research projects, including RCTs, from people with lived experience are growing,209,213,214 with concrete positives being documented.215 It is only by having this kind of foundation built together with EBEs that research projects can make a ‘real difference’ to the standards and diversity of care offered to people with mental health problems and to health services more broadly.216 It should also be noted that a proportion of the coinvestigators had lived experience of mental health problems and suicidal experiences which is considered to be a substantive asset.217,218 That said, one aspect which we would have liked to improve was a systematic examination and evaluation of the differences that working with the CARMers made to specific research processes. However, this is a project in its own right.
The third strength was that the number of eligible participants who, once informed about CARMS, no longer wished to continue was small. This pattern continued, in that retention rates at the 6- and 12-month follow-up time points were 75% or better and appeared comparable across the treatment and control groups. However, assuming that similar attrition rates across trial groups mitigates potential biases is somewhat naive.219,220 Hence, being able to follow up potential participants who no longer wish to continue with psychological intervention studies to determine why this was the case is important but challenging.221 From the data that we were able to collect, the main reason for discontinuing participation in CARMS was logistical. For instance, we were unable to progress with some participants because they moved area, left the UK, changed their contact details without informing their care teams and/or were living somewhat chaotically in different residences.
A convergent qualitative, quantitative and mixed-methods approach was followed throughout CARMS, with qualitative work being nested in the different phases of the RCT, which constitutes a fourth key strength. This strength produced tangible advantages evident when assessing acceptability of the suicide-focused therapy (CBSPp), examining mechanistic pathways and investigating implementational challenges and facilitators. This convergent methodological approach exemplifies best practice.222,223
A fifth strength was that a number of different ways of assessing the impact of suicide research on participants and also researchers and therapists were instituted. These included (1) directly exploring with participants their experiences of participating in suicide research115 (2) rigorous monitoring and classification of potential SAEs (3) following up all participants (with their permission) 24 hours after doing assessments and (4) developing, implementing and monitoring supervisory procedures with CARMS researchers and therapists, some of which were led by co-PIs, with others being peer-led. This is particularly important in any suicide research but was even more important when COVID-19 lockdown restrictions came into force and impacted CARMS.
A sixth and final strength relates to the flexibility of therapy.224 The therapy was manualised,1 which is a strength in its own right. However, within that manualised approach, a number of features afforded flexibility, including (1) the desire for a shared and collaborative approach to formulation,111 (2) relatedly, the centrality of the working alliance from the viewpoint of both therapists and participants which bolsters engagement and tends to impact favourably on outcome,225 (3) different delivery platforms being available, albeit dependent on COVID restrictions (e.g. in participant’s homes, on NHS premises, by phone, using online video) which can be not only practically advantageous but also empowering.196
Future research directions and recommendations
Future research directions are offered in the context that the treatment effect was not significant at month 6 for primary and secondary suicide outcomes.
- Freeman has recently demarcated 10 design principles to guide the development and assessment of psychological therapies155 which melds with our approach to theoretically founded therapy development.1 One principle is to identify and investigate key mechanisms underlying psychological problems, and a second related principle is to embrace the, often inherent, complexity of those mechanisms. The CARMS project illustrates both principles. That said, the work elucidated by CARMS is still just a starting point and somewhat exploratory. There is a need to (1) model in mechanistic pathways the effect of therapy process variables, such as therapy ‘dose’, level of formulation and the operation of the therapeutic alliance and (2) examine complex mechanisms using convergent qualitative, quantitative and mixed-methods designs.
- To take this a little further, we found evidence that appraisals of social support were amenable to change in the therapy group, which, in turn, reduced the severity of suicidal thoughts at month 6. Qualitative work identified that perceptions on continua of not mattering to mattering, and feeling disconnected to feeling connected were important from the perspectives of participants with lived experience of suicidal thoughts and acts. It is recommended that convergent methodologies are needed to build on this work using micro-longitudinal quantitative and qualitative approaches which can better understand how these transitions occur and how these transitions can be bolstered.
- Approaches which are grounded in realist evaluation methodologies226 examine theoretically framed multidomain and multilevel contextual factors which impact pathways from mechanism to outcome.227–229 Such approaches have the potential to identify and specify the contextual groups under which suicide-focused therapies work on a continuum from most optimally to least optimally which is important for future mechanism-oriented and implementation-oriented future research projects.
- At month 12, suicidal ideation severity was equivalent for both groups in the CARMS project. Examining ways of sustaining and amplifying reductions in suicidal experiences over time needs to be a priority. Promising approaches include peer-support online,230 in the community, on inpatient wards, and during the transition from hospital to the community. That said, development of peer-support has to be sensitive to the challenges accompanying specific mental health problems (e.g. suspiciousness, paranoia) and include careful and sustained training and supervision.
- One aspect of suicide prevention work, which is critical to further develop, is ways in which individuals living with severe mental health problems develop psychological resilience.231,232 Resilience work was out of scope and not feasible to conduct in the CARMS project, but is an important future research direction.
Clinical implications and recommendations
An initial broad point is that the clinical implications and recommendations that can be drawn from the findings of the CARMS project have applicability not only to a wide range of mental health professionals (e.g. psychiatrists, therapists, clinicians, nurses, social workers) but also to carers and families.
- There was no significant treatment effect at the 0.05 alpha level, meaning that adding a suicide-focused therapy, CBSPp, to TAU did not result in greater reductions in suicidal ideation severity compared to TAU alone. A clear implication is to better understand the extent to which different subgroups of participants benefited the most and the least from CBSPp, using for example, realist evaluation approaches.226
- The significant indirect (mediated) effect in which suicidal ideation reduced because appraisals of social support improved for those in the treatment relative to the control group provides a potential focus for suicide-focused therapy intervention. Nevertheless, the caveat that perceptions of social support, expanding to a sense of being meaningfully connected,46 can become complex in the context of both suicidal and psychotic experiences because both can act to inhibit social contact and denude and sense of meaningful engagement. Understanding how a robust and authentic therapeutic alliance can bolster a sense of meaningful connectedness is vital when working alongside people who have suicidal experiences.182,233,234
- CARMS qualitative work illustrated that a formulation-based suicide-focused cognitive talking therapy (CBSPp) was acceptable and feasible for a wide range of people experiencing severe and enduring psychosis and suicidal thoughts/behaviours.116,235 We found that CARMS participants, who were experiencing psychosis and who were extremely suicidal, wanted psychological therapy and wanted a suicide-focus in therapy.
- Our qualitative work also demonstrated how communicating about suicide, although wanted and needed, was sometimes perceived to be challenging by participants.46 This means that it is important to explicitly explore how different ways of communicating about suicide with professionals, carers, relatives and friends can be realised, such that emotional safety is tangible to clients.
- An unexpected consequence of working on CARMS throughout the pandemic was an observational finding that participants appeared not only able to tolerate working remotely but also actively embraced it. That said, we could not evaluate differential effects of mode of therapy delivery. Nevertheless, given the high levels of anxiety often experienced by people with psychotic experiences, offering remote forms of therapy may enhance engagement.
- Our suicide-focused therapy was delivered in one-to-one sessions by suitably qualified therapists. The high levels of formulation achieved by participants implied that they resonated with the suicide-focus of the work, and with the necessary commitment over a number of sessions (median = 16). Advances include exploring the effectiveness of group therapy, together with developing not only traditional face-to-face peer support together with online support forums230 but also peer delivery of suicide-focused therapy. Indeed, a wider point is the need to explore ways that non-mental health professionals can optimally work with people who live with suicidal experiences.