Evidence review
The systematic literature search identified three economic studies that assessed the cost effectiveness of specific psychological or psychosocial interventions compared with treatment as usual or routine care. All three studies were conducted in the UK (BYFORD1999 [Byford et al., 1999], BYFORD2003 [Byford et al., 2003], GREEN2011).
GREEN2011 was identified during an update search. The authors used a cost-effectiveness analysis, comparing group psychotherapy plus routine care with routine care alone for young people aged between 12 and 17 years who had at least two past episodes of self-harm within the previous 12 months. The analysis was conducted alongside an RCT in the northwest of England with a sample population of 181 for group therapy and 183 for routine care. The group psychotherapy comprised six weekly sessions initially followed by a booster of weekly sessions for as long as was needed, while routine care was made up of local children and young people's mental health services provided by CAMHS teams. The perspective of the analysis was societal with broad service use from the NHS, social services, education services, voluntary services and criminal justice services. The indirect cost due to productivity lost was tested in the sensitivity analysis. The primary outcome was proportion of participants who had not harmed themselves over the preceding 6 months at 12-month follow-up.
The reported total mean cost per young person (in 2005/06 prices) over the 12-month period was £21,761 (SD £38,794) for group therapy and £15,354 (SD £24,981) for routine care. No statistically significant difference in the two mean costs was detected. For the primary outcome result, the proportion of those young people who received group therapy and did not have any episode of self-harm over the follow-up period was 38.9% while that of the routine care arm was 41.9%. The reported incremental cost-effectiveness ratio (ICER) was £2,020 per 1% increase in the proportion of young people not self-harming, with the probability of group therapy being an optimal strategy ranging from 12 to 28% as willingness to pay for outcome improvement increased.
The application of the economic evidence of this study in the guideline is limited given that the perspective considered is societal and the final outcome was not measured in terms of quality of life values. Also, according to the authors, and from the estimated likelihood of the cost effectiveness of group therapy at increasing willingness-to-pay thresholds, the addition of group therapy to routine care is probably not more cost effective than routine care alone.
The second study by BYFORD2003 evaluated the cost effectiveness of MACT compared with treatment as usual for adult patients (16 to 65 years) with a history of self-harm recruited after presenting with an episode of self-harm. Their analysis was based on the clinical trial by Tyrer and colleagues (2003). Those requiring inpatient psychiatric treatment, or with psychotic or bipolar disorder or alcohol or drug dependence were excluded. The MACT group was given a treatment manual each and offered up to seven sessions of cognitive therapy while those in the treatment as usual group were offered standard treatment, which varied between the three studies and included problem-solving, psychotherapy, GP or voluntary group referral, and short-term counselling. A societal perspective was adopted for the analysis. Resource-use items included hospital and community health services, social services, voluntary sector services, community accommodation, criminal justice system, and participants' living expenses and productivity losses. The primary outcome measure used in the analysis was the proportion of participants who experienced a repeat episode of self-harm during 12-month follow-up. Quality of life years (QALYs) were also measured, by calculating European Quality of Life – 5 Dimensions (EQ-5D) utility scores, taken at baseline, 6 and 12 months.
The total mean cost over 12 months was £13,454 in the MACT group and £14,288 in the treatment as usual group (1999/2000 prices). The reported percentage of participants experiencing a repeat episode of self-harm over the 12-month period of follow-up was 7% lower in the MACT group whilst QALYs were 0.0118 lower in the MACT group. Taking treatment as usual as the base case, the reported ICERs when compared with MACT were -£120 per 1% reduction in percentage of participants with a repeat episode of self-harm (thus MACT was the dominant strategy) and £66,000 per QALY gained. Cost-effectiveness acceptability curves (CEACs) showed that MACT had more than a 90% probability of being cost-effective when using the percentage of repeat episodes of self-harm as an outcome. With QALYs as an outcome, MACT has higher probability of being cost effective at a threshold less than or equal to £66,000 per QALY. However, at different threshold values, the probability of MACT being more cost effective ranges between 44 and 88%. Extrapolating approximately from the CEACs, the probability of MACT being cost effective at a willingness-to-pay threshold of £20,000 and £30,000 were 65 and 60%, respectively.
The results of this study are highly applicable to this guideline in terms of the population, healthcare system, interventions and outcomes considered. However, the broader perspective, other than NHS and personal social services (PSS), taken by the study may be relevant to the population resource use but not recommended by NICE (2009d). Other limitations with the study findings were that uncertainty around the effectiveness measures were not presented. Given the small differences between the two treatment groups in terms of QALYs and percentage of repeat self-harm episodes, it is possible that these differences were not statistically significant and may explain why differences in percentage of repeat episodes but not QALYs favoured the MACT group. In addition, as noted by the authors, the chance that any coping mechanism could possibly improve quality of life may be plausible. In other words, with self-harm as a coping mechanism, such interventions that result in least reduction in repeat episodes of self-harm may be associated with more gain in quality of life than other interventions with significant reduction in repeat episodes of self-harm. Consequently, this calls for more caution in the interpretation of the direction of QALYs gained or lost with respect to self-harm interventions.
The third study, by BYFORD1999, evaluated the cost effectiveness of a home-based social work intervention plus routine care compared with routine care for children and young people (age range 10 to 16 years) who had self-poisoned. The home-based social work intervention delivered by two psychiatric social workers consisted of an assessment session and four intensive sessions targeted towards intra-familial communication, behavioural techniques and problem-solving. Routine care involved the visitation of psychiatric and psychiatric nurses in the clinic on an outpatient basis. The analysis was based on an RCT of 6 months' follow-up with outcome measures and costs reported for 162 children (77 for routine care and 85 for the intervention group).The perspective of the analysis included the NHS, PSS as well as the educational and voluntary sector. Resource use included assessment sessions, hospital services (inpatient, day-patient, intensive care unit, outpatient care and A&E services), GP visits, school nurses and doctors, community psychiatric nurses, counsellors, educational welfare officers, educational psychologists, social workers and foster and residential care.
The primary outcome measures used in the study were the Suicidal Ideation Questionnaire and BHS both of which were completed by the individual, and the Family Assessment Device (a measure of family functioning) completed separately by both the young person and their parents. No statistically significant differences were detected in any of the primary outcomes at 6 months between the two treatment groups. Similarly, no statistically significant differences in costs between the intervention and routine care (£1,455 versus £1,751; p = 0.6) were detected.
Regarding the applicability of this study to the guideline, it has a number of methodological limitations although the participant population, interventions and healthcare system considered in the study are all relevant. Firstly, there was no synthesis of incremental costs and outcomes or use of the QALY as a final outcome measure. Secondly, the short time horizon may not have allowed for full evaluation of all the important costs and effects associated with the intervention. Finally, the uncertainties of the result estimates were not tested.
Details on the methods used for the systematic search of the economic literature are described in Section 3.6.1. Information on the methods used and the results reported in the economic studies included in the systematic literature review are presented in the form of evidence tables in Appendix 14.
Economic modelling methods
Intervention considered in the analysis
The economic analysis considered interventions that were shown to be effective in reducing the number of repeated self-harm episodes according to the systematic review and meta-analysis of the clinical evidence. For the purposes of the economic model, the GDG identified a more realistic psychological intervention for reducing repetition of self-harm episodes to consist of six sessions delivered by a skilled and competent mental health worker with each session lasting for 60 minutes while the treatment as usual was described as consisting of a basic treatment provided by the CMHT to service users who self-harm after the initial hospital management of any associated acute physical and/or mental health problem. For the group receiving psychological interventions, treatment as usual is considered as a baseline intervention with the psychological intervention serving as an additional intervention. The psychological intervention is delivered either at the service user's home or in a clinic.
Model structure
The model structure construct aims to elaborate the natural history of the self-harm population as much as possible. Identified literature on the risk of repetition of self-harm showed a varied self-harm repeat risk ranging from 15 to 33% (Lilley et al., 2008b; Owens et al., 2002; Zahl & Hawton, 2004). Given the insufficient data on self-harm mortality and quality-of-life outcomes, a simple decision tree incorporating Markov nodes (represented by ‘M’ in , and Markov health states (self-harm and no self-harm) with an annual cycle length was constructed using Excel workbook 2007 to partly capture the treatment effects and costs of psychological interventions over a period of time in the future. According to the model structure, 1000 hypothetical cohorts of people aged 8 years and above who self-harm were provided with a psychological intervention plus treatment as usual or treatment as usual alone. People in each cohort either self-harmed after treatment, or were prevented from self-harming with ‘no self-harm’ taken as the absorbing state (see ). In the base-case analysis, the time horizon was taken to be 12 months based on the meta-analysis of the treatment effect of the psychological intervention lasting up to 12 months. A longer time horizon of up to 24 months was tested in the sensitivity analysis assuming the treatment effect was sustained till the end of the second year. A schematic diagram of the decision tree is provided in .
Costs and outcomes considered in the analysis
The economic analysis adopted the perspective of the NHS and PSS, as recommended by NICE (2009d) and reported in 2010 prices. Costs consisted of intervention costs (psychological intervention) and annual costs of care of a person who self-harms. The cost of treatment as usual was not considered in the analysis, as this was common to both arms of the model. The measure of outcome was the number of people prevented from a repeat episode of self-harm.
Clinical input parameters
Clinical input parameters consisted of the RR of repetition of self-harm associated with provision of a psychological intervention plus treatment as usual compared with treatment as usual alone, and annual baseline risk of repetition of self-harm following treatment as usual. Data were derived from the guideline systematic review and meta-analysis of clinical evidence. The baseline risk of repetition of self-harm estimated by Lilley and colleagues (2008b) to be 33% was found to be comparable with the pooled self-harm repetition risk from nine studies included in the meta-analysis. Given the possibility that the baseline risk of self-harm repetition used in the base-case analaysis can be an overestimation of risk of repetition of self-harm, as some people may not be presenting to services, a lower risk of repetition of self-harm is tested in the sensitivity analysis. In the base-case analysis, the economic model used the outcome measure assessed at last follow-up period (12 months on average) as agreed by the GDG.
Cost data
Cost of psychosocial intervention: The cost of intervention was estimated based on the descriptions of resource use identified from the psychological intervention studies included in the systematic review, confirmed by the GDG to be consistent with clinical practice in the UK. The intervention is a brief psychological therapy consisting of six sessions, which is provided by a nurse (mental health) specialist with each session lasting for 60 minutes. To estimate the intervention cost, the unit cost of a nurse specialist per hour of client contact reported in Curtis (2010) as £91 per hour was used. Calculation of this unit cost was based on the median full-time equivalent basic salary for Agenda for Change Band 6 of the January to March 2010 NHS staff earnings for qualified nurses. The estimation also included the salary oncosts, qualification costs, overheads and capital overheads (Curtis, 2010). Adjustment was made for those interventions provided at the person's home by adding the cost of travel time to hourly cost of client contact. The total mean cost of the psychological intervention was then estimated as the average cost of both home-based and non-home-based psychological interventions by multiplying the quantity of resource use by the respective unit costs (see ).
Summary of the average costs of psychosocial intervention.
Cost of self-harm: The estimation of costs incurred by a service user following an episode of self-harm was based on a retrospective cost analysis by Sinclair and colleagues (2010a), conducted in the UK. This study estimated costs following an episode of self-harm from the perspective of the NHS and PSS with a mean follow-up period of 10.9 years, which was divided into 6-month cost intervals. Among the 150 participants recruited into the cost study, 78 service users with available resource use in each time period were analysed. Resources measured in the study included primary care services, emergency department services, hospital services such as medical and surgical inpatient bed days, outpatient consultations, laboratory investigations, and inpatient psychiatric care. Other resources included were outpatient psychiatric care, psychotropic prescriptions, social service visits and social service residential placements. The cost estimate was reported as cost per episode of self-harm per 6-month interval and was £2,994 in 2004/05 prices. This estimate was inflated to 2010 price year using Hospital and Community Health Services pay and price inflator (Curtis, 2010) and also doubled to approximately estimate the annual cost of care for a person who self-harms to reach £6,998. According to GDG opinion, the cost incurred by people prevented from future episodes of self-harm after receiving a psychological intervention or treatment as usual was assumed to be negligible. provides the details of the clinical and cost input parameters described above with their probability distributions. For costs beyond 12 months, cost adjustment using a discount rate of 3.5% was applied as recommended by NICE (NICE, 2009d).
Summary of the base-case input parameters of the economic model.
Data analysis and presentation of the results
In the base-case analysis, the cost effectiveness of a psychological intervention plus treatment as usual versus treatment as usual alone at the 12-month time horizon was evaluated. Subsequently, evaluation at a longer time horizon of up to 24 months was made in the sensitivity analysis assuming the treatment effect was sustained till the end of the second year. Sensitivity analaysis was also conducted for other key parameters of the model. Two methods were employed to analyse the input parameter data and present the results of the economic analysis.
First, a deterministic analysis was undertaken, where data were analysed as point estimates; results are presented as mean total costs and outcomes associated with each intervention. Subsequently, an ICER was calculated, expressing the additional cost per additional unit of benefit associated with one intervention relative to its comparator. Estimation of such a ratio allows consideration of whether the additional benefit is worth the additional cost when choosing one treatment option over another. Alternatively, if one intervention is less costly and more effective than its comparator, then this is obviously the most cost-effective option (dominant) and no ICER needs to be calculated.
To test the robustness of the results under different scenarios, one-way and two-way sensitivity analyses were conducted. The following scenarios were explored:
A resource-intensive scenario comprising 12 psychological intervention session delivered by a Band 7 clinical psychologist was considered to reflect the possible variations in resource inputs and the associated ICER.
A scenario of 50% variability in the cost of self-harm from
Sinclair and colleagues (2010a) was tested to examine the effect on the ICER level since the reported cost of the self-harm estimate has a wide standard deviation around the mean cost.
It was assumed that people prevented from self-harming following a psychological intervention incurred a negligible future cost. The possibility that these people may incur some cost such as four subsequent GP visits (per clinic consultation lasting 17.2 minutes is £53 inclusive of direct care cost [
Curtis, 2010]) was tested to examine the implication of such extra cost on the ICER level.
A lower baseline risk of repetition of self-harm of 24% following an index episode was tested given that the baseline risk used in the base case was based on those presenting to services with the possibility of overestimating the risk as some individuals will not present to service. From the literature, the annual risk of repetition of self-harm varies from 15 to 33% (
Lilley et al., 2008b;
Owens et al., 2002;
Zahl & Hawton, 2004) with 24% as an approximate estimate of the average annual risk of repetition of self-harm.
Variations in the effectiveness of the psychological intervention using the upper and lower values of the 95% confidence interval of the relative risk and baseline risk of repetition of self-harm episodes was also tested.
In addition to deterministic analysis, a probabilistic analysis was also conducted. In this case, all model input parameters were assigned probability distributions (rather than being expressed as point estimates), to reflect the uncertainty characterising the available clinical and cost data. Subsequently, 10,000 iterations were performed, each drawing random values out of the distributions fitted onto the model input parameters. This exercise provided more accurate estimates of mean costs and benefits for each intervention assessed (average results from the 10,000 iterations), by capturing the non-linearity characterising the economic model structure (Briggs et al., 2006). The distributions assigned to each of the input parameters are shown in .
Results of probabilistic analysis are presented as mean costs and effects derived from 10,000 iterations, as well as in the form of CEACs, which demonstrate the probability of each intervention being cost effective at different levels of willingness-to-pay per unit of effectiveness (that is, at different cost-effectiveness thresholds the decision-maker may set).
Results of economic modelling
For clarity, the results the analyses were presented as follows:
For the base-case analysis, the probabilistic and deterministic estimates showing the mean cost and mean effect of both psychological and treatment as usual strategies with the resultant ICER evaluated at the end of the last follow-up period (that is, 12 months) (see ).
The sensitivity analysis showing the value of ICERs for a given parameter(s) of interest evaluated both at the end of the last follow-up period and 24-month time horizon.
Base-case analysis at 12-month time horizon.
shows the mean costs and number of people prevented from self-harm repetition for each of the interventions assessed in the analysis. The ICER evaluated as £ per additional person prevented from repetition of self-harm episode was £164 (probabilistic analysis). The ICER estimate by deterministic analysis was £46. The ICER estimate from the probabilistic analysis is relatively higher than the deterministic estimate and is regarded as more reliable given the variations around the parameter inputs.
Discussion
The economic analysis undertaken examined the cost effectiveness of a psychological intervention as an additional intervention to treatment as usual compared with treatment as usual alone. The result of the economic modelling showed that to prevent an additional person from repeating an episode of self-harm by choosing six sessions of a psychological intervention delivered by a nurse specialist instead of treatment as usual alone, the NHS will be incurring an additional cost of approximately £200. Also demonstrated by this analysis was that the psychological intervention plus treatment as usual has a greater likelihood of being cost effective compared with treatment as usual alone at various willingness–to-pay levels of £1000 and above. Hence, choosing the psychological intervention will depend more on the service provider's level of willingness-to-pay for an additional person prevented from self-harm repetition.
An important point in this analysis is the resources used to deliver the psychological intervention. The model analysed a realistic option as suggested by the GDG comprising six sessions delivered by a skilled and competent nurse. Nevertheless, some of the reviewed studies described up to 12 sessions delivered by a clinical psychologist. Though the benefit of extra sessions and service delivery by a clinical psychologist could not be ascertained from the reviewed studies, it may be worth examining further to identify the advantages and/or possible disadvantages of such an intensive option. However, from the sensitivity analysis in , such an intensive option may be incurring a much higher cost compared with the realistic option should the benefit of the two options be similar. Also, when the long-term resource impact of implementing a psychological intervention is considered, the model shows that psychological intervention has the potential to be a more cost-effective option notwithstanding significant uncertainty around some parameters.
Limitations of the analysis
The major issue that may limit the usefulness of this analysis is the non-availability of QALYs estimates. Nevertheless, from the reported potential gain of more QALYs following treatment as usual compared with fewer QALYs gained following MACT in the study by BYFORD2003, it is uncertain whether QALY gain or loss is a useful measure of outcome in long-term self-harm management. In the same study, the authors were of the opinion that self-harm as a coping mechanism may be associated with improvement in quality of life than other measures used to prevent it.
Another limitation is variation in the modalities of psychological interventions among the studies included in the meta-analysis. Though the probabilistic method used in this analysis substantially accounts for the associated uncertainties, it is important to interpret the result of this analysis with caution especially in relation to the cost of intervention and the RR.
