This is the first comprehensive systematic review of TW interventions for people with LTCs. There are many various forms of TW interventions that have been suggested as potential therapies for the treatment of mental and physical illnesses. However, in this systematic review, unfacilitated EW such as described in 1986 by Pennebaker and Beall1 was the most frequently used approach. As few as five studies66–70 reported a facilitated type of TW intervention. However, even these five studies are not that similar to the form of TW which has been described by TW practitioners as the most frequently used writing therapy in both NHS and voluntary sectors in the UK.
Statement of main findings
Systematic review
Five studies66–70 were identified which examined the effect of facilitated TW on four different disease areas (two studies in PTSD,69,70 and one each in dementia,67 psychiatric problems,68 and serious physical problems – primarily cancer66). The actual TW interventions studied were extremely varied, two being internet based. Three of the studies66,67,70 were very small, and across the studies quality was mostly unclear. The studies measured different outcomes but all reported positive outcomes in favour of the writing intervention, suggesting that facilitated TW may be beneficial but it requires much more evaluation, which should be conducted rigorously.
Unfacilitated EW was evaluated in 59 studies.71–119 There was considerable heterogeneity in participants, interventions (context, content, delivery mode), comparators (factual writing, time management, non-writing, waiting list or SMC) and outcomes (type measured and length of follow-up). For measurement of outcomes, 172 instruments were used and more than 300 different measures were reported. All but four52,53,78,87 of the studies were RCTs. Methodological quality was frequently unclear, but most studies were likely to have introduced performance or detection biases that could have increased the likelihood of finding a positive result.
Economic considerations
The review of economic evidence did not identify any true economic evaluations, quantifying both the monetary costs and health consequences of TW interventions. Only one of the studies89 included in the systematic review attempted to quantify the cost of delivering the TW intervention itself and made an estimate of cost savings. Some studies did report on the impact of TW on patients’ use of other health services: there was a significant reduction in reported use of medications with TW, but no significant effect on the use of other health services. Costs for delivering programmes of TW are likely to vary, depending principally on the amount of paid practitioner time per participant. The estimated direct cost to the NHS ranged from as little as £12 for unfacilitated interventions to > £2000 for a series of 16 one-to-one facilitated TW sessions. There is insufficient evidence to estimate how much, if any, of this cost would be offset by savings from reduced use of other NHS services. There is also insufficient evidence to estimate the overall cost-effectiveness of TW to the NHS. Cost–consequence analysis suggested that there might possibly be a favourable balance of participant benefits to NHS costs for selected interventions in selected LTC groups.
Realist synthesis
The realist synthesis was able to develop two programme theories that (to some extent) explained some of the findings within the studies included in the systematic review. Uncertainty continues to exist over why individuals with LTCs might wish to participate in TW (of any variety) and what, how, and in what way these individuals hope to gain from TW. For unfacilitated EW, there remains uncertainty as to how and why it generates any outcomes of interest, for whom, and in what circumstances. In facilitated TW the relative importance of the different processes within it (i.e. what processes need to be used – how and why – to generate desired outcomes) is unclear.
In unfacilitated EW, when participants were offered the option of undertaking EW, this formed the starting context. When offered this resource, participants made a choice about whether or not to participate. The participant reasoning in response to the resource offered might be threefold: (1) an initial assessment as to whether or not the individual felt or thought there was any value in participating (i.e. will he/she gain any benefit); (2) was this a safe environment in which to write; and (3) was there any support on offer to them? These three responses formed the most prominent mechanisms and their contextual triggers are discussed in more detail with reference to facilitated TW. Studies using unfacilitated EW generally did not attempt to create a safe writing environment or provide support. Once a participant had decided to undertake unfacilitated EW, they were then in a position to experience for themselves its effects – which may range from harms to benefits. No benefit or perceptions of harm appeared to result in non-engagement or inconsistent engagement. When the individual perceived benefits, he/she seemed to engage in further TW. This perceived benefit seemed to accumulate with repeated engagement and became a new contextual trigger for the mechanism related to perceptions of benefit. The range of benefits that a participant obtained from TW appeared to be individually specific; they may have been immediate or delayed to some degree, and not necessarily predictable in advance or those that were measured by the researchers in the included studies.
The programme theory for facilitated TW is an elaboration of that for unfacilitated EW. Within the NHS and voluntary sector, TW is usually offered as an activity within a group context. Potential participants either self-refer or are encouraged to attend. These groups tend to operate in a context in which actual participation in TW is voluntary and non-participation tolerated. Such contexts enables potential participants to try out (the mechanism here being trialability) the group – seeing for themselves how it works and what goes on. A range of group related contexts appear to be important in influencing potential participants to participate with a TW technique. With repeated attendance at the group, potential participants can subjectively and objectively assess the (contextual influences of) degree of tolerance, support and guidance (from others and the facilitator), the range of activities on offer, voluntariness, interactivity of the group, and so on. These appear to trigger the mechanisms of perceptions of benefits, safety and support. Potential participants feel safe to take part in TW, when they perceive that they are likely to receive support from the group if potentially troubling emotions arise. If they perceive subjective benefit, continuing participation is more likely to occur. The programme theory for facilitated TW and unfacilitated EW are the same once a participant has decided to undertake TW. Within the realist synthesis we were unable to clarify further the mechanisms of how different TW techniques are meant to produce desired outcomes.
Strengths
Systematic review
The systematic reviews and realist synthesis were supported during all phases by the expert advice of four TW professional practitioners. The systematic review has been undertaken following PRISMA criteria63 and was registered with PROSPERO at the start of the project. Comprehensive database searches were adopted, supplemented by checking citation lists, theses and grey literature and contacting experts in the field.
Selection criteria were developed to include only participants with documented diagnoses of LTCs and the list of LTCs considered was generous, with careful consideration of whether or not specific medical conditions could be considered long term (e.g. patients with bladder papilloma following resection). No restrictions were applied regarding the intervention (i.e. all types of therapies falling within the TW umbrella were considered). Included was any comparative study, not only RCTs, and careful consideration of appropriate comparators was given; where the control intervention appeared to have an inadvertent EW component, for example when patients with cancer were asked to write about the facts of their treatment, these studies were not included.
The assessment was not restricted to stated primary outcomes only. Independent double data extraction for all numerical data was used. The reporting of all results is very comprehensive and analyses were very thorough. Study authors (32 out of 64 included studies) were contacted for missing numerical data and 14 responded. In the report results were categorised by ICD-10 code as a systematic way of dealing with the very wide range of conditions in which TW had been investigated. Categorising by ICD-10 code is an unusual way to present systematic review results, and can be used sensibly only when an intervention can be applied to a wide range of conditions – in this case both psychiatric and physical conditions. Meta-analyses were conducted both within and across conditions to fully explore any potential effects of these interventions.
Economic considerations
The review of published economic evidence was nested within a thorough systematic review, following accepted methodological guidelines. An attempt was made to estimate the costs of TW from a NHS perspective, based on information about the range of interventions provided in the included studies and expert advice on how TW is used in practice. The cost–consequence summaries brought together best-available evidence of effects on outcomes of importance to patients with simple estimates of NHS costs for three illustrative case studies.
Realist synthesis
This realist synthesis has been undertaken following (where applicable) the RAMESES (Realist and Meta-review Evidence Synthesis: Evolving Standards) quality standards for realist syntheses.146 Two programme theories have been developed. Document selection and appraisal applied the principles of relevance and rigour, and data extraction has focused on extracting data that are capable of supporting programme theory refinement. Where possible, a realist logic of analysis has been applied to the data and the synthesis has been reported following the RAMESES publication standards for realist syntheses.146 Reporting for the realist synthesis has been distributed across different sections of this report as some of the processes used are similar to those of the systematic review. For example, the searching process of the systematic review was used as the starting point of the realist synthesis. This also meant that the document flow diagram is shared across both reviews. Additional searching (of a limited nature) was undertaken and this has not been shown in the main document flow diagram (see ). Based on the focus of the realist synthesis (to explain the outcome patterns found in the systematic review) no scoping of the literature was undertaken. No changes were made to the initially planned review process other than a much more limited additional searching due to time constraints. The RAMESES publication standards have been followed and an overview of compliance may be found in Appendix 7.
Weaknesses
Systematic review
We were not able to identify a UK-based therapeutic practitioner of unfacilitated EW as an expert advisor. We double checked only 10% of our retrieved titles and abstracts, so there is a risk that we missed relevant studies; however, our 10% checks did show excellent agreement between reviewers. We did not exclude studies in which the underlying condition was poorly described (e.g. patients with various cancers). The facilitated/unfacilitated TW split was a post hoc decision based on the background literature21 and discussion with the TW expert practitioners collaborating in the project. We tried to be consistent with our decisions around selection of interventions and comparators, particularly in the meta-analyses, but not all interventions and comparators were similar, introducing some heterogeneity. For example, when a study had two interventions – one with positive writing and one with standard unfacilitated EW – the latter group was selected. If a study had one intervention only, it would be included in the meta-analyses even if participants were asked to write about positive aspects of their experience, particularly on the last day of writing.
A disadvantage of using ICD-10 codes to categorise studies meant we had a large number of categories, often with relatively few studies in them, and this, in turn, constrained the sensitivity analyses that we could conduct within these categories.
Studies not reporting any numerical data (raw scores or proportions for any one of the outcomes) were excluded from the systematic review. Therefore, relevant studies might have been omitted. However, strenuous efforts were made to contact study authors regarding missing data. Authors were not contacted about study quality or when an explanation of a given instrument was not provided.
Many of the meta-analyses had high levels of statistical heterogeneity and should be interpreted with caution. Additionally, meta-analyses are on small numbers only. A post hoc decision was made to conduct overall meta-analyses across all LTCs for the outcomes of depression and anxiety. This decision was made before knowing the individual LTC meta-analyses results and was based on the large number of studies reporting these outcomes, and their clinical usefulness as an exploratory analysis.
Regarding the study designs, both RCTs and non-RCTs were included but no sensitivity analyses of randomisation compared with non-randomisation design were performed. This was because there were so few studies within each ICD-10 code that further subdivision was thought to be unhelpful. We also did not conduct any sensitivity analyses by study quality; we described study quality comprehensively but did not use it further in our analyses.
We made a pragmatic decision not to investigate subgroup analyses within the included studies because almost all appeared to be post hoc analyses, and the studies were not sufficiently powered for these analyses. Therefore, we did not look for moderators, which was in the aims and objectives in our original proposal. Regarding interventions, self-completion books were included as a TW intervention.
Economic considerations
The lack of evidence on cost and cost-effectiveness from published comparative studies was a major challenge. Most studies did not report impact on health-care resource use, only 12 out of 64 studies included such information.57,68,75,77,83,89,90,97,98,104,109,119 In addition, the reporting standards for resource use in most papers were weak. It was also difficult to estimate the cost of the interventions in some studies, owing to a lack of detail on the level of practitioner input. The diversity of the studied populations, the inconsistent and sparse nature of the available effectiveness evidence, and, particularly, the lack of evidence on QoL outcomes, made de novo decision modelling an unrealistic prospect.
Realist synthesis
Although searches were exhaustive, selection criteria for the realist synthesis could have been identified earlier in order to conduct even more iterative searches. There was only one qualitative study147 identified as being useful for the realist synthesis process. Additional searching would have been desirable as it would have enabled more relevant data sources (especially those that contained potentially relevant existing substantive theories) to have been identified. However, because of time constraints, only one single round of additional searching specifically for the realist synthesis was undertaken. The absence of these additional searches is likely to have impacted on the degree of the explanatory powers of the realist synthesis. In addition, it was not possible to include the studies found in the updated search (to January 2015) in the realist synthesis.
The selection criteria for the realist review could have been more closely defined, or tailored earlier, in order to capture more relevant literature. Studies investigating healthy students and non-comparative studies were identified as a potentially useful tool for further programme theory development.
The development of the programme theory relied predominantly on the data contained within randomised, non-randomised and case–control study data. Previous review teams have identified that the data necessary for conducting a high-quality realist synthesis is often not found in RCTs. However, as the included studies consisted of a broader range of study types and much larger numbers of studies than that found in Dieleman et al.148 and Kane et al.149 it was anticipated during the project that sufficient data would be available. To supplement the data within the studies included in the effectiveness review, two additional searches were undertaken: one to look for studies on facilitated TW and the other for studies of any type that reported on one or more of the included studies. Despite these two additional searches, gaps in the data existed, which meant that it was not possible to fully elucidate and test a number of aspects of the programme theories, such as:
The mechanisms underlying the TW techniques encountered in this project. Additional searching for data from different disciplines (e.g. humanities) might have been more informative. In addition, there exists a large body of research on the use of unfacilitated EW in populations without LTCs and data from this body of work might have been valuable in mechanism identification and elucidation.
The role of the group in facilitated TW requires greater elucidation. The realist synthesis identified a number of contextual influences and mechanisms, but these are unlikely to be the only relevant ones. A large body of work exists on groups in general and on therapeutic groups specifically. Further searching in these areas would likely have provided further relevant data. The role and competencies of the group facilitator also needs further elucidation and additional searching for data on this topic area would have been helpful.
The value placed by participants in TW on the presence of an audience was noted in some of the data. This is another area with a large body of literature located within disciplines outside of health services research that is worthy of further attention.
This is the first realist synthesis that has sought to develop a realist programme theory that attempts to explain the outcome patterns for unfacilitated EW and facilitated TW when used for LTCs. In other words, it is the first attempt that tries to unpack the black box of TW when used in a population with LTCs. The programme theory provides potentially transferable knowledge about how these types of TW techniques might fare under different contexts. However, an important caveat on this programme theory is that it requires further development and refinement. This is because it includes only some of the relevant and important mechanisms and related contextual triggers. The paucity of relevant data in the studies used to develop and refine the programme theory has had an impact on both its detail and extent. In addition, the strength of the inferences made has been affected by the rigour of the included studies. Many of the outcomes reported within the included studies must be considered to be less trustworthy as a result of the conduct of the studies. For example, all reported findings of benefits, harms, correlations and/or associations were treated with caution when used in programme theory development. Findings that could reasonably be considered to be more trustworthy were those that had not been as a result of statistical manipulation and/or interpretation by the authors. Consideration was still given as to how such data were generated, for example the interview method for qualitative data. The result is that many of the inferences made in the development of the programme theory were often tentative. The implication is that further testing and refinement of the programme theory is needed, through future secondary or primary research.
This project faced two significant methodological challenges in trying to undertake both a systematic review and realist synthesis. The first related to the purpose of the realist synthesis and the other to human resources required. The purpose of the realist synthesis was to explain the outcome patterns reported in the studies included in the systematic review. Thus the realist synthesis was not able to proceed in earnest until it had been clearly decided what the final set of included studies were for the systematic review. This process for confirming the final set of included studies for the systematic review did not occur until month 11/12 of the project. This meant that the realist synthesis could only begin in the last third of the project. An additional related challenge that emerged from the focus of the realist synthesis was around the nature of the included studies. About midway through the project, the expert advisors had indicated that within the NHS and voluntary sector for LTCs, the main form of TW used was facilitated TW. Only when the final set of included studies for the systematic review had been finalised and were being analysed did it emerge that the vast majority of these related to the use of unfacilitated EW. This unexpected finding meant that additional unanticipated work was needed in the realist synthesis, as it had to develop one or more programme theories to make sense of both unfacilitated EW and facilitated TW.
In terms of human resource requirements, it must be kept in mind that both systematic reviews and realist syntheses are labour intensive. Within this project the majority of the researchers time was devoted to the systematic review. This meant that one of the two reviews had to be truncated – in this case the realist synthesis. Once the included studies had been identified, data extraction, analysis and synthesis took place in parallel for the systematic review and realist synthesis. As the project had only one full-time researcher (OPN) and almost all of her time was devoted to the systematic review, less human resources allocation, and time was available for the realist synthesis before the expected end date of the project. Within the 3-month time frame left to undertake the realist synthesis (especially in light of the additional work needed, as explained above), it was not possible to undertake the additional searches that would be needed to more fully develop the programme theories for unfacilitated EW and facilitated TW. These would have been especially important, as the studies included in the systematic review did not contain sufficient relevant data. There was simply not the spare human resources capacity or time to do so.
The challenges mentioned above provide important lessons to review teams and funders who wish to undertake both combined systematic reviews and realist syntheses.
Sequencing If the purpose for undertaking a realist synthesis is to explain the outcome patterns of studies included in a systematic review then additional time is likely to be needed to enable the rigorous execution of the realist synthesis component. Identifying the included studies takes time and the realist synthesis can pragmatically start only once it is clear what the included studies are. Its start is thus delayed and it is likely that an additional minimum of 9 months (ideally 12–18 months) is needed for a rigorous realist synthesis to take place. Review teams and funders need to expect that such combined reviews are going to take longer to complete.
Human resources Both systematic reviews and realist syntheses are labour intensive. Planning and budgeting needs to reflect the need for additional researcher(s) time when a project gets to the stage when data extraction, analysis and synthesis is taking place for both the systematic review and realist synthesis. Otherwise, one or the other of the reviews is likely to suffer. Ideally, two different individuals with sufficient funded time are needed with the necessary skill sets to undertake the systematic review and realist synthesis separately.
Uncertainties
Systematic review
Most interventions evaluated were unfacilitated EW and did not mirror those currently used by professional TW practitioners in clinical practice in the UK. The main uncertainty is the clinical effectiveness of TW as it is practised by therapeutic practitioners, including within the NHS. It must be noted that the facilitated TW interventions included in the systematic review do not mirror those used in current NHS practice. There is insufficient clinically relevant evidence on facilitated TW as practised within these studies to know whether or not it is beneficial. It is uncertain if unfacilitated EW might be harmful, particularly in psychotic patients, but there was little evidence of harm in the studies evaluated.
Economic considerations
The economic case for the NHS to fund TW interventions for people with LTCs remains unproven. Depending on the level of practitioner input, these interventions are likely to be relatively inexpensive, and there is some evidence suggestive of possible cost savings related to reduced use of other health care, although this evidence is sparse and inconsistent. The cost–consequence case studies also suggest that there may be positive impacts on measures of health and well-being for some types of TW in some patient groups but these are exploratory analyses. Until there is robust evidence of patient benefit, it is difficult to conclude that TW would be a cost-effective use of NHS money.
Realist synthesis
Unfacilitated TW appears to have some effect in that the people who took part did notice some changes to intermediate or more proximal outcomes. However, the outcome of interest was health functions of some form, but TW does not seem to have much of an effect on this final desired outcome.
Within another realist synthesis, one could try to work out what outcomes (final desired and/or intermediate/proximal) might be influenced by unfacilitated TW. This would be the programme theory development aspect of a realist synthesis of unfacilitated TW. In this project we could touch on this only because of insufficient relevant data. This further realist synthesis could try to build a more detailed realist programme theory [i.e. one that further explains what has caused each outcome (final/intermediate) and under what contexts], the causal force being the mechanism(s).
We did not look at unfacilitated EW on healthy students but these studies may or may not inform what outcomes (final/intermediate) might be possible with unfacilitated TW, under what contexts and what the mechanisms might be. So there may be contexts within which it is reasonable to extrapolate what can be learnt from the healthy students studies to those who are unhealthy – based on making (and if possible testing) the assumption that the same mechanism(s) might be in operation under these different contexts.
The programme theory for unfacilitated EW and facilitated TW in LTCs highlights how complex these interventions are. It cannot and should not be assumed that they will work as expected when applied in differing contexts. The outcomes participants will get from TW are at present neither predictable nor necessarily achieved. In addition, these outcomes are rarely the ones hoped for by the researchers who have undertaken the primary studies included in the effectiveness reviews and realist synthesis. A number of contexts appear to influence whether or not individuals will choose to participate and continue with TW. In the real-world setting of clinical practice (such as in the NHS and voluntary sectors), due consideration needs to be given to ensuring that these contexts exist (or can be made to) if there is to be any hope that meaningful and continued participation in TW is to take place. Of note is that many of these contextual influences are beyond the TW technique itself and sit in the wider world that surrounds the individual and the TW technique. Despite the development of the programme theory, at present there is too much uncertainty surrounding too many aspects of it to enable a firm recommendation to be made about the benefits for the large-scale introduction of unfacilitated EW or facilitated TW into routine clinical practice.
