Introduction
In a health-care system such as the NHS it is important to consider not just the benefits and potential harms of interventions for the recipients, but also the impact that interventions may have on the use of limited health-care resources. In the previous chapter the available evidence on the effects of TW on health outcomes for people with a variety of LTCs is summarised. Here the evidence on the economic impacts of TW in these populations from a NHS perspective is reviewed. NHS provision of TW interventions would incur direct costs; the cost for staff training and time to deliver the intervention, or of payments to external experts; administrative costs; and costs for the use of a room and, possibly, some materials. There may also be some indirect costs or savings for the NHS if TW encouraged patients to increase or decrease their use of other health services. For example, patients might feel better able to manage their condition and consult their GP or hospital specialist less often or, conversely, TW might lead patients to recognise a need to make more use of available health services. The net effect on NHS expenditure might therefore be positive or negative.
In addition to financial costs and savings, a full economic evaluation would account for the intrinsic value of any impacts on patients’ health and well-being. For example, health gains or losses attributable to the intervention could be quantified in terms of quality-adjusted life-years (QALYs), where one QALY is defined as 1 year lived in perfect health for one person. The cost-effectiveness of the intervention could then be summarised as the additional cost per QALY gained, in relation to some appropriate comparator. For such a calculation, the magnitude and persistence of any effects of TW on HRQoL would need to be estimated. This calculation of QALYs from the studies in this review is problematic because although some of the studies reviewed in the preceding chapter reported on general HRQoL (23 out of 64 studies), only one study88 used a measure suitable for QALY estimation, such as the EQ-5D, and none used the Short Form questionnaire-6 Dimensions (SF-6D).
From the early stages of the project, the need for flexibility in the approach to the consideration of the economic evidence was apparent. Given the limitations of the evidence base, it was concluded that it was not possible to directly estimate QALY impacts of TW across the range of populations with LTCs. Consideration was given to the possibility of using a decision-analytic model to estimate the effects of the intervention on intermediate indicators of disease progression, and then to link to the effects of disease progression on health-care costs and outcomes. However, the lack of convincing or consistent evidence of such effects led us to conclude that such modelling exercises would not be appropriate or feasible. Instead, a more pragmatic approach was taken, making the most of the evidence available.
The economics section is in three parts:
a systematic review of TW studies reporting on economic outcomes (resource use, costs and/or cost-effectiveness)
an estimation of the cost of providing a range of TW interventions in a NHS context
case studies presenting balance sheets of available economic and clinical evidence for three conditions: PTSD, RA and breast cancer.
Systematic review of therapeutic writing studies with resource-use outcomes
Methods
A review across LTCs of the available literature on comparative studies of TW for patients with LTCs reporting economic outcomes was conducted. This resource-use systematic review was nested within the overall systematic review described in the previous chapter and the realist synthesis presented in the following chapter. The full search strategy is given in Appendix 3.
The inclusion criteria used for the economic systematic review are shown in .
Inclusion criteria for resource-use systematic review
Titles and abstracts were checked by two reviewers to find relevant studies, data extraction was undertaken by one reviewer and checked by a second. If there were disagreements, a third senior systematic reviewer or health economist was consulted to make a decision. All of the studies in the economic review were also included in the effectiveness review, and critically appraised as part of that review (see Appendices 4 and 5). Quality assessment using a critical appraisal checklist for economic evaluations was planned if any economic evaluations were identified.
Where three or more studies presented outcomes in the same category (e.g. health centre visits, medications used), meta-analysis was conducted using the same methods as for the effectiveness systematic review.
Results of economic review (resource use)
No full economic evaluations (cost-effectiveness, cost–benefit or cost–utility studies) were identified. One study89 estimated that the cost of a writing intervention was US$130 per patient, based on a psychologist’s fee and some administration time. This study89 found a reduction in inpatient use for the EW intervention compared with SMC and estimated that this represented a cost saving of US$25,878 per patient per year. Another study113 estimated that the cost of delivering videotaped instructions for an EW intervention was approximately US$5 per patient. The basis for this estimate was unclear but it seems to be reasonable.
Twelve studies57,68,75,77,83,89,90,97,98,104,109,119 that reported on some element of health-care resource use were identified. One study reported on the cost of the TW intervention89 and none on monetary estimates of costs or savings relating to differences in health-care use between the study groups. Consequently, no additional economic quality assessment was conducted.
Study details
Details of the studies57,68,75,77,83,89,90,97,98,104,109,119 are outlined in . The publication dates ranged from 1996 to 2012: six studies57,75,77,90,104,109 were published after 2009. The majority of the studies57,77,83,89,90,97,109,119 were from the USA. Two studies75,104 were conducted in the UK. All patients were recruited via flyers distributed at disease-specific clinics or approached by health-care professionals during their treatment cycle. The disease areas varied, but can be broadly grouped into three categories: cancer (n = 3),75,77,83 chronic pain or FM (n = 3),57,109,119 or PTSD, mental health disorders or drug dependency (n = 4).68,90,97,98 The remaining studies related to MI104 and cystic fibrosis.89 One study68 was a case–control study and tested a facilitated form of TW – an internet discussion. The other studies57,75,77,83,89,90,97,98,104,109,119 were RCTs and tested unfacilitated EW.1 The control interventions varied: seven studies68,75,83,89,97,98,109 included a normal care or non-writing control arm. Six studies57,77,89,90,104,119 used a non-emotional form of writing.
Details of studies reporting resource use
Numerical results of resource-use studies
Numerical results are shown in . The quality and detail of studies reporting on health-care resource use varied widely across studies. The types of use recorded were broadly similar: almost all reported on contacts with health-care services, consultations with clinicians or inpatient stays (n = 11). Four papers57,68,83,104 also reported on use of medication. Willmott et al.104 also reported the number of weeks’ absence from work as an outcome, which is an important indicator for the personal and broader economic impact of an intervention.
Contact with health-care services
Of the 11 studies57,68,75,77,83,89,90,98,104,109,119 reporting on some form of contact with health-care professionals in a health-care setting, seven studies83,89,90,98,104,109,119 provided means and/or SDs that could be combined in a forest plot (). Golkaramnay et al.68 and Mosher et al.77 reported proportions of patients accessing services. The meta-analysis results showed no significant differences (SMD –0.19, 95% CI –0.57 to 0.18). Willmott et al.104 reported on two outcomes relating to health-care contact: median number of GP and hospital visits, which was higher in the control arm, and the mean number of cardiac rehabilitation sessions attended, which was higher in the intervention group. Overall, the results suggest that TW has little impact on health centre visits.
Forest plot of health-care resource use. df, degrees of freedom; IV, inverse variance.
Use of medication
Three studies57,83,104 reported means and SDs for the impact of TW on use of medication, and Golkaramnay et al.68 reported the proportion of participants receiving medication. Stark57 reported on the number of units of pain and psychotropic medication per patient per month, finding lower levels of use in the intervention group 10 weeks after baseline. Willmott et al.104 and Rosenberg et al.83 took a different approach, looking at the number of medications participants report using at similar time spans post baseline (5 and 6 months, respectively). In both cases, participants in the intervention group were identified as using fewer medications, although in Willmott et al.104 this difference was not significant, and Rosenberg et al.83 did not report on significance. Willmott et al.104 also noted a potential time effect – medication use decreased in the intervention group but increased in the control group. The meta-analysis results () suggested that fewer medications were taken after TW (SMD –0.28, 95% CI –0.54 to –0.02).
Forest plot of medication use. df, degrees of freedom; IV, inverse variance.
Many of the costs associated with the use of medication are indirect, and not examined in these studies, for example consultations with prescribers, and treatment for side effects. Although these add to the uncertainty around impact of decreased medication use, it also raises the possibility that there are potentially unacknowledged benefits associated with TW, which have not been explicitly accounted for here.
As with health-care contacts, this evidence base is not strong enough to draw conclusions about the likely impact of TW interventions on patient use of medication.
Costs of therapeutic writing interventions
None of the identified studies presented estimates of the cost of TW or EW in the UK. Therefore, estimates for a range of interventions in a NHS context were made, after discussion with practitioner experts.
The primary driver of costs is likely to be the staff costs associated with employing (or contracting) TW practitioners. TW practitioners come from a wide range of professional backgrounds, and many are self-employed, rather than NHS employees. However, for illustrative purposes, the costs for NHS occupational therapists published by the Personal Social Services Research Unit (PSSRU)126 were adapted ().
Unit cost of practitioner time (adapted from PSSRU 2013)
Costs were estimated for NHS staff at Agenda for Change pay bands 5, 6, 7 and 8a to reflect a range of levels of qualifications and experience. The estimated total cost per hour ranged from £36 to £63 (at 2012–13 prices). For comparison, the PSSRU report126 estimated a cost of £36 per hour for a hospital-based occupational therapist; £48 per hour for a counsellor in primary care; and £59 for a clinical psychologist.
All of the above costs were inclusive of staff training, as well as salary on-costs (employer’s national insurance plus 14% of salary for employer’s contribution to superannuation); management and administration overheads, and capital overheads (annuitised over 60 years at a discount rate of 3.5%). The capital cost included an allocation for patient care and non-patient care facilities within a NHS hospital, and can therefore be assumed to include the cost of a room for delivery of the individual or group TW sessions. There may be some other costs for materials, such as printed instructions, videos, pens and paper, but these are difficult to estimate.
The time required for a TW practitioner to prepare for a session was not quantified. The PSSRU report126 estimated the percentage of working time spent on face-to-face client contact: 77% for primary care counsellors and 45% for clinical psychologists (no estimates were available for occupational therapists). This implies that, for every hour of client contact, approximately 15–30 minutes will be spent on other activities. For cost estimates, it was assumed that it takes 10 minutes for a TW practitioner to prepare for a session with clients; this is not assumed to vary according to the length of the session or the number of participants in the session.
The majority of studies included in the systematic review used self-administered unfacilitated form of writing. This typically involves a brief initial consultation with the patient, either in person or over the telephone. Following this initial discussion, the patient is left to continue with the intervention on their own, writing unsupervised, often in their own home, for the prescribed time and on the prescribed topic. There is minimal subsequent interaction with practitioners – in some cases, the TW is not returned to, or read by, practitioners. Although this set of methods is used in the majority of studies – possibly due to its relative ease of implementation – this form of writing is not representative of current clinical practice in the UK. Practitioner experts advised that facilitated writing is much more common in a NHS setting: writing sessions run by trained experts. These sessions may be individual or group, and patients may attend multiple sessions.
The estimated costs for a range of TW interventions, with varying group sizes, numbers of sessions and contact times per session, are outlined in .
Illustrative costs for a range of TW interventions
The upper and lower limits of these ranges are based on interventions described in the included studies in Chapter 3. Session length for facilitated TW ranged from 45 minutes in Lange et al.69 to 120 minutes in Rickett et al.66 and Sloan et al.70 The length of the intervention varied from four sessions in Sloan et al.70 to 16 sessions in Hong and Choi.67 The number of participants in the interventions ranged from two to eight, based on advice from a TW practitioner that sessions range from one to seven participants, with four being the optimal number. It was assumed that the introductory session for unfacilitated writing would last for 10–30 minutes. Even the more resource-intensive, facilitated versions of TW seem to be low cost in relation to other health-care interventions. However, the range of estimated costs is wide: from £17 to about £2200 per patient. The estimated cost per patient for unfacilitated EW here is less than the US$130 estimated in Taylor et al.89 In that study89 the cost was largely based on the psychologist’s fee of US$130 per hour, which is more than the £59 per hour for a UK clinical psychologist quoted above.
Without adequate evidence on the effects of TW on patients’ health and well-being, or its effect on other health-care expenditure, it is difficult to draw conclusions about whether or not TW is a cost-effective use of NHS resources.
Exploratory cost–consequence analyses
Methods
Studies included in the effectiveness review spanned a wide range of disease areas and populations, making it difficult to identify where TW may have value. Three disease areas were examined in more detail, bringing together effectiveness evidence alongside estimates of economic impacts (cost–consequence analyses). The areas chosen for further analysis were PTSD, inflammatory arthritis and breast cancer. These topics were chosen by the SGC before the results of the meta-analyses were known, with the aim of reflecting a range of different conditions for which TW has been used. There was also a pragmatic element to this choice – these topics contained a greater number of studies, including some of the higher-quality ones. It was felt that they were likely to provide a stronger evidence base than some other areas. It is important to note that these case studies are unlikely to be representative of the whole evidence base for TW.
For each case study, a summary of effectiveness evidence was prepared, and presented alongside estimates of costs and use of health-care resources. Where possible, costs were estimated using information about the intensity of practitioner input in the related clinical studies.
Case study 1: post-traumatic stress disorder
In total, four studies69,70,98,121 reported on the use of TW in the treatment of people with PTSD. Gidron et al.98 and Smyth and Arigo121 tested unfacilitated TW to treat PTSD. Details of these studies are presented in Chapter 3 (pp. 78–83). Lange et al.69 and Sloan et al.70 evaluated a facilitated TW intervention in a PTSD population (see Chapter 3, F43: post-traumatic stress disorder).
Summary of study design and quality
Two studies69,70 evaluated individual forms of facilitated TW compared with waiting list controls (see Table 5). Both recruited from the community: Lange et al.69 recruited online and screened for post-traumatic stress and grief; those in the Sloan et al.70 study were recruited through local advertisement and had a primary diagnosis of PTSD related to a MVA. Study quality was mixed (see ). The Lange et al. study69 was at risk of bias because of non-reporting of randomisation, concealment of allocation or blinding of outcome assessment. Sloan et al.70 was a randomised trial, but did not report on blinding of participants or outcome collection.
The unfacilitated studies compared EW with factual or time-management writing as control in patients with diagnosed PTSD (see Table 32). Smyth and Arigo121 was conducted in the USA, and Gidron et al.98 in Israel. Both studies had methodological and reporting flaws that left them susceptible to bias (see ).
Estimated costs of intervention
The two studies69,70 of facilitated writing gave quite detailed information about the treatment protocol and therapist input, which was used for costing. Lange et al.70 evaluated a 5-week internet programme (Interapy), consisting of 10 45-minute writing sessions. Feedback on submitted writing was sent to each patient by a therapist on seven occasions. The feedback consisted of about 450 words. The time taken by the therapists to prepare this feedback was not reported but is unlikely to have been much less than 1 hour. The mean basic salary for qualified clinical psychologists is £46,280, similar to Agenda for Change band 8a.126 However, the therapists employed in this study were graduate and postgraduate students in clinical psychology (mean age 29 years), not yet fully qualified, but who had attended advanced courses in behavioural cognitive psychotherapy and received special training in using writing assignments in PTSD. For costing purposes we assumed a band 6 salary (£30,712), or £45 per hour including indirect costs and overheads (see ). The estimated cost per participant is therefore in the region of £315 (7 × £45).
Sloan et al.70 evaluated a written exposure therapy (WET) intervention consisting of five weekly sessions (one lasting 60 minutes and four lasting 40 minutes). Participants had individual contact with a clinician for approximately 25 minutes during the first session, and for 10 minutes in each of the remaining sessions. In addition, it was assumed that therapists would need some time to prepare for sessions: 10 minutes for the first session and 5 minutes for each remaining sessions. Total therapist time per participant is therefore approximately 95 minutes. Therapists were clinicians with masters or doctoral level qualifications and prior experience of treating PTSD with exposure-based therapies. Therefore, it was assumed that a band 8a salary, with a mean cost per hour of £63, was reasonable. The estimated cost participant is in the region of £100 (95/60 × £63).
It is more difficult to estimate the cost of the unfacilitated writing interventions. Participants were asked to write on three or four occasions, for 15–30 minutes per session. However, after an initial briefing, patients wrote on their own, either at home or in a health-care environment. Neither study of unfacilitated TW in PTSD reported who instructed participants on the writing tasks, or how long this took. A cost of between £12 and £42 for unfacilitated writing interventions was assumed (see ).
Summary of costs and consequences
presents an overall summary of evidence relating to the clinical effectiveness and cost of TW for people with PTSD. It can be seen that evidence for facilitated TW is sparse but promising: the included studies reported significant benefits for the intervention group compared with control, for a range of outcomes of importance to patients: PTSD symptoms, measures of emotion, anxiety, depression, sleep and somatisation. The estimated cost of delivering these interventions in a NHS context is relatively modest: in the region of £100–300. However, evidence about the impact of facilitated TW on patients’ overall QoL and well-being, and on their use of other health-care services is lacking.
Summary of evidence on TW costs and consequences of unfacilitated EW in PTSD
Evidence relating to unfacilitated forms of TW is sparse and inconsistent. Of the two included studies,98,107 one reported some positive effects on mood and a biomarker for stress (Smyth et al.107). However, the other study98 reported negative effects on PTSD symptoms and somatisation and an increase in non-routine health visits.
Case study 2: inflammatory arthropathy
Six studies107,113–117 reported on the use of TW in the treatment of people with inflammatory arthropathy and five studies107,113–115,117 were used for this analysis (i.e. except Lumley et al.’s116 which was found in the update searches). Details of the studies were presented Chapter 3.
Summary of study design and quality
All five studies107,113–115,117 evaluated an unfacilitated EW intervention compared with neutral writing (time-management control). In addition, Broderick et al.113 included a second intervention group, who were asked to write about the meaning of the trauma, and Lumley et al.’s115 included a second control group, who were asked to write about a positive emotional event. All studies recruited patients with diagnosed RA, except Hamilton-West and Quine114 which recruited participants with a diagnosis of AS. Follow-up ranged from 13 to 43 weeks, and outcomes included measures of disease activity, pain, function, mood depression and QoL. None of the studies reported on health-care use or costs. Wetherell et al.117 and Hamilton-West and Quine114 were conducted in the UK, and the other four studies107,113,115,116 were conducted in the USA. The quality of the studies was mixed (see ). Three were classified as true RCTs,107,114,115 but all had design or reporting flaws that left them susceptible to bias.
Estimated costs of intervention
As in PTSD, it is difficult to assess the cost of the interventions because papers did not generally report on the grade or qualifications of the staff who gave patients instructions on the writing task, or specify how long this took. The location of the writing intervention varied: in Smyth et al.107 participants wrote in a private room in a laboratory; Broderick et al.,113 Wetherell et al.117 and Lumley et al.115 adapted the intervention for participants to write in their home; Hamilton-West and Quine114 did not report the location of the intervention.
Broderick et al.113 produced a videotape to introduce the rationale for the writing task and to provide detailed instructions for patients. They noted that this method was chosen as there is evidence that patients respond well to video-based introductions, and because it was likely to be a cost-effective approach that could be reproduced across large numbers of patients with minimum input from professionals. However, Broderick et al.113 noted that some physician time would still be required to introduce the idea of the intervention to participants, and to encourage them to participate. In addition, a cost would be incurred for each patient given a video: including the cost of materials and reproduction; and a proportion of the cost for the development and production of the video. The latter is very difficult to estimate, as it is unknown how many patients would use the video.
In Wetherell et al.117 participants were contacted by telephone before and after each writing session. Assuming 10 minutes’ preparation, 10 minutes for the introductory conversation and 5 minutes per telephone call before and after each of four writing sessions, the total estimated practitioner time to deliver the intervention is approximately 60 minutes, incurring a cost in the region of £36–63 per patient, depending on the grade of the practitioner.
Summary of costs and consequences
Summary results for inflammatory arthropathy are presented in . In this case, the results look rather more promising for unfacilitated EW. The meta-analysis of measures of disease activity in the four RA studies107,114,115,117 found a significant benefit in the intervention group at short-term follow-up (8–10 weeks). Three of these studies114,115,117 also reported significant positive effects for some outcomes at some time points: measures of mood in Wetherell et al.117 pain in Lumley et al.115 and function in Hamilton-West and Quine.114 However, no significant effects were found across a number of other outcomes – including QoL – and Lumley et al.115 reported some negative effects on mood immediately after writing.
Summary of evidence on TW costs and consequences of unfacilitated EW in inflammatory arthropathies
Case study 3: breast cancer
Eight studies53,54,74–79 reported on the use of unfacilitated EW in participants with breast cancer; see Chapter 3 for details.
Summary of study design and quality
The studies evaluated unfacilitated EW against a usual care comparator53,54,74,75,78 or against an un-EW control74,76,77,79 (note that Craft et al.74 had two control groups). The participants were at various stages of disease and treatment: Jensen-Johansen et al.,76 Gellaitry et al.75 and Craft et al.74 recruited women with early-stage breast cancer who had recently completed treatment; Mosher et al.77 recruited women with metastatic breast cancer who were in significant distress. The studies were all conducted in the USA, with the exception of one Korean study78 and one Danish study.76 There were three randomised studies,54,74,79 but all studies were subject to design or reporting bias (see ).
Estimated costs of intervention
Henry et al.53 sent written instructions to participants by post, asking them to write for 20 minutes, on one occasion at home. The cost of this intervention would be minimal, say £5 for materials. Three studies74,75,79 reported a single face-to-face or telephone contact with participants, either before the first writing session74,79 or after the final session.75 A cost of £12 for these three interventions74,75,79 was assumed. Two studies76,77 reported a rather more resource-intensive approach, with telephone calls to the participants before and after each writing session. Jensen-Johansen et al.76 first contacted women participating in another study by mail. The intervention consisted of three 20-minute writing sessions at home over a 3-week period. Participants were telephoned by a research assistant, trained by a clinical psychologist, before and after each writing session. Assuming 10 minutes’ preparation, 10 minutes for the initial telephone call and 5 minutes for each subsequent call, the total contact time would be 45 minutes: costing £27 (at £36 per hour). Mosher et al.77 included women with advanced disease and high levels of distress. Women were screened and recruited by telephone, and then sent written instructions, paper and envelopes to return their writing. The intervention consisted of four writing sessions over 4/7 weeks, each lasting from 20 to 40 minutes. The women were telephoned by a psychology research fellow before each writing session, after 20 minutes, and, sometimes, again after an additional 20 minutes. The contact time was therefore greater than in other unfacilitated writing interventions. Assuming that the initial telephone call lasted for 20 minutes, and that each subsequent call lasted for an average of 5–10 minutes, the total contact time would have been approximately 60–90 minutes. At £36 per hour, the total cost per patient would be around £36–100.
Summary of costs and consequences
Summary results for breast cancer are presented in . The evidence base for unfacilitated EW in breast cancer is rather larger than in the other case studies presented above, with eight comparative studies,53,54,74–79 but the results are not encouraging. Although some studies did report some significant effects, meta-analyses of measures of positive mood, negative mood and depression at short-term follow-up failed to find a significant treatment effect.
Summary of evidence on TW costs and consequences of unfacilitated EW in breast cancer
