The Scottish Government

« Previous | Contents | Next »

Listen

ANNEX G: OUTCOMES OF THE SCOPING REVIEW

G.1 The aim of the scoping review was to provide a broad overview of study findings, irrespective of judgements regarding study quality. In taking this approach, the scoping review is able to provide an account of the literature in this field as a whole. It provides, for example, insight into which modes of intervention have been pursued in the literature to date, which population groups have been focussed on and which methodologies researchers have chosen to adopt in evaluating interventions for suicide and self-harm. It also provides an indication of which interventions show some promise of efficacy, albeit in the absence of the high quality trials which may still need to be carried to confirm possible benefits.

G.2 Finally, evaluating outcomes 'across the board' allows us to carry out multivariate analyses to explore which of a potential range of variables impacts to the greatest extent on the outcomes identified for an intervention. Is it, for example, the quality of the research methodology alone which determines whether an intervention is identified as being effective, or is it the population in which the intervention is evaluated, or the mode of intervention or the type of behaviour being addressed, or do all of these distinct aspects have an impact? Taking into account only the outcomes of the small number of highest quality studies would not allow these sorts of questions to be addressed and in a comparatively sparse empirical literature such as this, these are questions which are of considerable importance in driving forward the search for effective interventions.

Outcomes for suicide

G.3 Defining outcomes very broadly on the basis of an increase or decrease in the rate or absolute number of suicide(s) based on whichever method of accounting the study authors used, outcomes for the 67 studies addressing completed suicide were as follows:

G.4 The one study demonstrating an increase in suicide following intervention (Oerlinghausen et al 1994) was a retrospective follow-up study of psychiatric patients treated with lithium. Within those studies providing only a narrative report of positive outcomes (that is, presenting no quantitative data to back up the textual report of positive outcomes), only one followed a qualitative methodology (Landers 1981). Outcomes for this study are tenuous, since they rely on generalising from a single case study of carbon monoxide poisoning to conclude that changes in carbon monoxide legislation may have an impact on suicide. Epidemiological studies provide support for this view, but the study as it stands is best seen as an exploration of the mechanisms by which this is achieved (providing practitioners with a lengthier period of time in which to intervene).

G.5 Two other qualitative evaluations (Etzersdorfer 1993, Owens et al 2004) report, respectively, a case study of treatment for depression and a psychological autopsy evaluating detection and treatment of depression. Neither study reports successful outcomes. Two other studies, also providing purely narrative accounts of successful intervention were RCTs. The remainder followed other quantitative methodologies. An evident difficulty with outcomes based purely on qualitative methods is the problem of establishing objective parameters within which to confirm outcomes. When quantitative studies also fail to provide adequate statistical analysis there are additional grounds for scepticism regarding the reliability of assumed outcomes.

G.6 The comparative rarity of suicide accounts for the lack of statistical analysis in fewer than half of the studies following a quantitative methodology but failing to present supportive analyses. One study, (Oyama et al 2004), which evaluated a community-based intervention programme for older people living in rural areas, did provide some statistical analysis supporting the author's conclusions in respect of outcomes for particular subgroups at particular stages in the study. However, the overall statistics presented did not appear to support the broader conclusions given by the author in the study text. Taking these issues into consideration, we have reasonable grounds to believe that around one third of the studies report potentially meaningful reductions in suicide. Tables G1 and G2 overleaf provide a summary of study outcomes for all specific interventions evaluated for the prevention of suicide.

Table G.1 Outcomes for Interventions to Reduce Suicide: Pharmaceutical

Table G.2 Outcomes for Interventions to Reduce Suicide: Non-Pharmaceutical & Multi-Modal Interventions

G.7 The overview of interventions and outcomes provided in the above tables demonstrates that research effort evaluating interventions for the prevention of suicide has been spread too thinly over broadly distinct areas. Whilst around a third of studies report positive outcomes, support for individual interventions is diminished by the small number of studies addressing each distinct intervention and the equivocal nature of outcomes across the board. Only two approaches to prevention stand out as having a small but not negligible number of studies available, with largely unequivocal outcomes. These approaches are the restriction of access to means and maintaining ongoing contact with the suicidal person. Multi-modal approaches also seem promising, although on the basis of a slightly smaller number of studies and with the rider that the specific interventions combined within each study differ.

G.8 Approaches which have been evaluated in a number of studies with little evidence of efficacy include the training of health professionals (e.g. to recognise and treat depression) and a variety of institution-based and public-health oriented programmes. In the latter case however, the studies are again evaluating programmes which rarely contain the same components, so comparison across the full range of programmes may be misleading. The use of lithium, clozapine and 'anti-depressants' (taken as a composite of any anti-depressant drug given to participants) have all also been evaluated by a non-negligible number of studies. On balance, lithium appears the most promising of these pharmaceutical approaches. However, concerns must be raised in respect of one study (Oerlinghausen et al 1994) which reported an increase in suicide during lithium treatment. Notably given the fact that in one additional high quality study (Tondo et al 1998) increases in the risk of 'suicidal acts' compared to baseline were reported in the first year following discontinuation of lithium. The potential impact of clozapine and 'anti-depressants' evaluated as a composite is at best equivocal. Given the limited number of studies addressing each of the interventions listed and the diverse contexts in which interventions have been evaluated, positive outcomes must be treated with considerable caution, even as suggestions for promising avenues to pursue in further research into the prevention of suicide.

G.9 To explore further the pattern of outcomes, we compared the characteristics of studies reporting significant reductions in suicide with those reporting either neutral outcomes or increases in suicide. Studies with positive outcomes were marginally more likely to have used national statistics to establish suicide than other means (77% of studies with positive outcomes versus 51% with neutral or negative outcomes, ? ²=4.4 one-sided p<0.03). This is not a function of sample size (there was no significant association between methods of establishing outcomes and sample size within this group of studies), but is in line with expectations based on the nature of official statistics (cf. Douglas 1967; Maxwell Atkinson 1978). There were no significant differences depending on whether studies were retrospective or prospective or whether outcomes were measured at discrete time points or were evaluated on a continuous basis across a period of time. Nor were there any significant differences in outcome when RCT designs were compared with other quantitative approaches. Since only three qualitative studies were identified it was not possible to compare across broader methodological categories.

G.10 Little can be established regarding the impact of participant demographics on study outcomes, for reasons given earlier. Neither ethnicity nor gender is sufficiently well differentiated within this group of studies to evaluate outcomes along even these broad demographic divides. Since the bulk of studies fell within the 'adult' (26-65) age group specified by the Research Advisory Group, little can be gauged regarding the impact of age on outcomes based on the priority age groupings as defined earlier. However, there was a significant correlation between the mean age of participants included in studies and the likelihood of positive outcomes (r=0.49 p<0.01), with outcomes more likely to be positive in studies with an older mean age of participant. It is of note that very few studies evaluating interventions to prevent completed suicide focus on the youngest age groups.

G.11 Given the fairly small number of studies addressing suicide and the wide spread of origins and approaches, other characteristics of the studies had to be compressed into very broad categories to allow meaningful comparisons. On this basis, there were no differences in outcomes between studies published recently and in earlier decades (2000s vs 1990s and pre-1990s combined); studies from different countries ( US and Canada vs all other populations), studies addressing different modes of intervention (pharmaceutical vs non-pharmaceutical) or studies addressing different populations (general population vs psychiatric populations vs other populations). Considering only those studies which both started and ended in the same setting, there was a marginal difference between studies where intervention took place in the community and where it took place in more controlled settings, with outcomes for more controlled settings showing a greater likelihood of success (? ²=3.63 one-sided p<0.05), as would be anticipated on the basis of research addressing the control of other-directed aggressive behaviour (cf. Leitner et al 2006). Finally, whilst quantitative studies demonstrating positive outcomes were likely to be of higher methodological quality ²³ than studies which failed to cite positive outcomes, this difference failed to reach significance.

Outcomes for attempted suicide

G.12 A total of 74 studies addressed attempted suicide as an outcome. Four of these studies failed to give sufficient details of this particular outcome to judge whether the intervention evaluated had been successful or not. Outcomes for the remainder of the studies are given below:

G.13 The one study identifying an increase in attempted suicide (Leon et al 1999) was a prospective follow-up of fluoxetine treatment in patients with affective disorder. As with the earlier profile of studies addressing completed suicide, it is worth noting that only three of the ten studies which report a decrease in attempted suicide without reference to confirmatory statistical analysis are qualitative studies. These qualitative studies evaluated the use of naltrexone (Krupitsky et al 2001), DBT (Perseius et al 2003) and training for school personnel (Ross 1980) using single and multiple case studies and survey designs respectively. One additional qualitative study (Kuipers & Lancaster 2000) reported a reduction in the likelihood of attempted suicide following the provision of informal support to brain injured patients. Of the quantitative studies which, as was the case for completed suicide, form of the bulk of evidence available for interventions in attempted suicide, 41% are RCTs.

G.14 Tables G.3 and G.4 overleaf give a summary of outcomes for all specific interventions evaluated for the prevention of attempted suicide. The available research is spread over an even more diverse range of interventions than was the case for completed suicide and with the possible exception of treatment with clozapine, which is compared only against other active comparators, no interventions stand out as having a reasonable number of studies reporting unequivocally positive outcomes. Treatment with lithium again offers some promise on the basis of outcomes from three of five studies, but concerns regarding possible increases in suicidal behaviour for at least some sub-groups again argue for caution in the use of this treatment.

G.15 The number of studies addressing school-based intervention programmes (11) is substantially higher than for competed suicide, and five of these studies report decreases in attempted suicide as a consequence of the intervention, but set against this are six studies with no evidence of any reduction in repetition and which on the whole also use stronger measures of outcome (actual behaviour rather than scale-based report). Again it should be noted that the programmes themselves are not directly comparable. Amongst the psychotherapeutically oriented therapies, DBT appears most promising on the basis of the available evidence. As with completed suicide, 'anti-depressants' evaluated as a group find little support in the intervention literature, with all five trials (including two trials comparing anti-depressants with placebo) reporting no reduction in suicide attempts.

Table G.3 Outcomes for Interventions to Reduce Attempted Suicide: Pharmaceutical

Table G.4 Outcomes for Interventions to Reduce Attempted Suicide: Non-Pharmaceutical & Multi-Modal Interventions

G.16 Outcomes for attempted suicide, using the broad categories which were available for analysis, seem even less sensitive to characteristics of study design, intervention, setting and population than is the case for completed suicide. We found no association between the use of RCTs and other quantitative designs and positive outcomes (again the number of qualitative studies was too small to include in the analysis). Similarly outcomes were insensitive to whether study designs were prospective or retrospective and whether outcomes were measured using discrete or continuous follow-up. The recency of studies also had no impact, with studies published in the 2000s no more or less likely to record positive outcomes than studies published one or more decades ago.

G.17 There were also no differences between the modes of intervention we were able to compare (pharmaceutical vs other interventions); how outcomes had been established (formal hospital or other records vs self-report vs other means) or between the different populations and settings considered. As previously, the only meaningful demographic comparison which could be drawn was based on the mean age of participants in the sample and in this case there was no association between age and the likelihood of a positive study outcome. The sole study characteristic showing even marginal association with the likelihood of positive outcomes for attempted suicide was country of origin, with US and Canadian studies less likely to report a reduction in suicide attempts as a consequence of intervention in comparison to all other countries (42% versus 64% ? ²=3.7 p<0.04). However, in contrast to outcomes for completed suicide, study outcomes were associated with study quality. Positive outcomes were significantly more likely to be reported by studies meeting at least the median quality score for their design group (77 % of studies with positive outcomes fell into this category versus 53% falling below the 'cut-off' score assigned for quality, ? ²=4.54 p<0.04). This suggests that future improvements in design quality could potentially help in the identification of additional interventions with positive ouctomes.

Outcomes for self-harm

G.18 The number of studies evaluating outcomes for self-harm were substantially fewer than for either completed suicide or attempted suicide (N=44, 22% of all studies). The distribution of outcomes across studies was as follows:

G.19 The two studies reporting statistically significant increases in self-harm following an intervention (Hopko et al 2003, Martinez et al 2005) both report on pharmaceutical interventions. The first study is an RCT comparing treatment with mianserin to treatment with either placebo or nomifeserine for people with borderline personality disorder, the significant increase in self harm was noted in those treated with mianserin. The second study reported on a case-control evaluation of SSRIs versus either tricyclic or other anti-depressants for people receiving their first prescription of anti-depressants. The significant increase in self-harm was noted in those treated with SSRIs.

G.20 As with both completed suicide and attempted suicide, narrative reports of outcomes were not wholly restricted to qualitative studies. Two quantitative studies of self-harm (one an RCT) also reported positive outcomes without supporting evidence from statistical analysis. The number of qualitative studies available for analysis was marginally higher in the case of self-harm (N=6) than in completed suicide or attempted suicide but again provided little opportunity for in-depth analysis of promising interventions, since only one of these studies (Perseius et al 2003ps) reported positive outcomes. This study, which is also referred to above in the context of attempted suicide, reported ten case histories of DBT for borderline personality disorder, with outcomes limited to self-report by the patients that therapy had reduced their self-harm. Of the quantitative studies, 67% (12 studies) followed an RCT design. Tables G.5 and G.6 overleaf summarise outcomes for all included studies evaluating outcomes for self-harm.

Table G.5 Outcomes for Interventions to Reduce Self-Harm: Pharmaceutical

Table G.6 Outcomes for Interventions to Reduce Self-Harm: Non-Pharmaceutical & Multi-Modal Interventions

G.21 A number of points can be made regarding the above tables. The most striking is the evident focus on psychotherapeutic rather than pharmaceutical intervention. This is in sharp contrast with the distribution of studies for completed suicide and attempted suicide. Since there are far fewer studies in total addressing self-harm and yet the range of interventions evaluated remains equally wide, the available evidence for any given intervention in this context is even more severely limited than in the context of completed or attempted suicide. The single most promising intervention taking the studies overall, without regard to quality or other considerations is DBT. All other interventions identified in the included studies have been evaluated by one, or at most two, studies. Other distinctions between the profile of studies for self-harm and the profile for suicide and attempted suicide are the lack of broad-based public health or educational interventions evaluated. Finally, although this is evidenced by only two studies, it is worth noting that behaviour therapy has been evaluated (with successful outcomes) in the context of self-harm in the studies we have identified, but has not been evaluated in the context of other forms of suicidal behaviour or of suicidal ideation.

G.22 In comparing across study characteristics, the profile for self-harm broadly matched that of completed suicide and attempted suicide, with positive outcomes significantly associated with very few study or participant characteristics. Study design did impact to some extent (again there were too few qualitative studies to include these in the equation) in that studies reporting positive outcomes were more likely to follow an RCT design than any other quantitative method (53% versus 45% ? ²=6.38 p<0.04). Other design characteristics, such as prospective versus retrospective follow-up or continuous versus discrete assessment of outcomes, had no significant impact. Date of publication, country of origin, population, setting and mode of intervention also had no impact on outcomes.

G.23 Demographic variables were again too poorly reported to draw comparisons between 'successful' and 'unsuccessful' interventions. However, studies with positive outcomes were marginally more likely to include participants recruited from the 25-65 age group than participants recruited from either the youngest or oldest age categories (73% versus 41% ? ²=4.05 one-sided p<0.04). This is a characteristic of evaluations commonly reported in intervention research (cf. Hahn et al 2005). The greater likelihood that RCTs would report positive outcomes was also in this instance reflected in the marginally greater likelihood that studies reporting positive outcomes reached at least median quality criteria for their design category (73% of studies with positive outcomes versus 41% of studies not reporting positive outcomes, ? ²=4.05 one-sided p<0.04). Again this suggests that improvements in study quality may help in identifying additional interventions with some promise of efficacy.

Outcomes for suicidal ideation

G.24 Nearly half of the studies we identified for inclusion in the review (N=94, 47%) evaluated outcomes for suicidal ideation. Three studies failed to clarify whether outcomes were positive or not. For the remainder, the distribution of outcomes was as follows:

G.25 The one study identifying a significant increase in suicidal ideation following an intervention (Cunningham-Owens et al 2001) was an RCT focussed on a community-based educational intervention for people with schizophrenia. The majority of the studies using quantitative designs (57%) also followed an RCT methodology. Studies using suicidal ideation as an outcome measure accounted for half of the qualitative designs (N=13, 50%) included in the review. In contrast to outcomes for the suicidal behaviours outlined above, this provides a reasonable number of qualitative studies (N=11) with positive outcomes to analyse.

G.26 As with the qualitative studies referred to earlier (two of which also report outcomes in this context), the quality and depth of the material available is, however, disappointing. Six of the studies were case studies, one (Perseius et al 2003) presented what could be regarded as a brief content analysis, the remainder followed no specific qualitative methodology, instead providing a narrative overview of survey or interview responses, relying on small numbers of participants, or on available audit and other data only peripherally linked to suicidal ideation. The interventions and populations addressed were also too diverse to allow for any comparative analysis across the qualitative studies.

G.27 In summary, the nature of many of the studies available does not provide firm support for the interventions evaluated. Taken at face value, positive outcomes were reported for the following interventions: occupational therapy, naltrexone; intervention by psychiatrists or by friends and family members in preference to intervention by GPs for high risk groups; anti-depressant treatment for people with major depression; short-term hospitalization for 'non-compliant' patients, DBT for people with borderline personality disorder, psychoanalytic psychotherapy and community or school-based programmes for adolescents and young adults and Magnetic Field Therapy for depressed patients with Multiple Sclerosis. These and other study outcomes for all interventions evaluated in the context of suicidal ideation are set out in Tables G.7 and G.8 overleaf. It is important to note that the positive outcomes reported by study authors in the context of these studies can be regarded as preliminary outcomes at best. Before they can be regarded as in any way definitive, further evaluation using more robust methodologies is required.

Table G.7 Outcomes for Interventions to Reduce Suicidal Ideation: Pharmaceutical

Table G.8 Outcomes for Interventions to Reduce Suicidal Ideation: Non-Pharmaceutical & Multi-Modal Interventions

G.28 The high proportion of studies addressing suicidal ideation is matched by the diverse range of interventions evaluated. Although around half of the studies (N=45) focussed on suicidal ideation alone, without reference to suicidal behaviours, the nature of the interventions evaluated broadly matched the profile for suicidal behaviours as outlined above. The range of individual pharmaceutical interventions evaluated was more diverse than that noted in connection with suicidal behaviours, but again focussed, predictably, primarily on various forms of anti-depressant. None of the pharmaceutical interventions evaluated have a sufficient number of studies, in particular high quality studies, demonstrating unambiguously positive outcomes to make any confident assertion regarding their likely efficacy. The most promising outcomes, with all available studies reporting reductions in suicidal ideation following the intervention, are for fluvoxamine (two studies) and sertraline (three studies).

G.29 The one form of non-pharmaceutical intervention evaluated by a substantive number of studies (N=13, school based intervention programmes) shows considerable ambiguity in outcomes. Again, this may be due to the diversity of distinct programme components addressed, but nevertheless this fails to provide evidence-based grounds on which to recommend strongly such programmes. Aside from the restriction of access to means, which again is supported by both studies evaluating its efficacy, all of the non-pharmaceutical interventions evaluated by more than a single trial show the same degree of ambiguity.

G.30 The age categories specified by the Research Advisory Group were better represented in the context of studies for suicidal ideation than for suicidal behaviour. However, no significant association between either these age categories or the mean age of study participants and reported outcomes was apparent. Other demographic characteristics were again too poorly reported to allow any analysis of possible associations between participant characteristics and study outcomes. Background characteristics of the study, specifically date of publication, country of origin of participants and how suicidal ideation had been established (narrative self-report vs scale-based self-report vs other means) also failed to show any significant association with the likelihood of positive outcomes. Study design did impact to some extent, with, as previously, positive outcomes more likely to be reported by RCTs than by other quantitative designs (56% versus 41% ? ²=7.89 p<0.02). Qualitative studies were, as previously, excluded from the design based analysis due to small numbers. No association was shown between outcomes and whether follow-up was continuous or discrete, prospective or retrospective.

G.31 In contrast to suicidal behaviour, the mode of intervention did have a significant impact on outcomes for suicidal ideation. Studies evaluating pharmaceutical interventions were significantly less likely to report positive outcomes than studies evaluating non-pharmaceutical interventions (36% versus 64.1% ? ²=4.08 p<0.04). Also in contrast to suicidal behaviour, outcomes were population-dependent. Studies focussed on psychiatric populations were more likely to report positive outcomes than studies focussed on other populations (61% versus 38% ? ²=7.49 p<0.01). Study quality also had a significant impact on outcomes for quantitative studies, independently of whether designs were RCT or non- RCT. Positive outcomes were far more likely to be reported by studies achieving at least the median quality score for their design category (82% versus 18% ? ²=4.69 p<0.05). The consistency of this finding across different modes of behaviour is promising, in that it supports the view that improvements in study design may have the potential to allow identification of additional interventions with positive outcomes.

Comparing across the types of suicidal behaviour

G.32 Outcomes for completed suicide and, to a lesser extent, for suicidal ideation were evaluated in isolation from outcomes for other suicidal behaviours by a substantial proportion of studies (N=40 and 45 respectively). In contrast, outcomes for self-harm and attempted suicide tended to be addressed in tandem either with each other or with completed suicide or suicidal ideation. Only 11 and 13 studies respectively focussed solely on outcomes for either self-harm or attempted suicide and presented these separately rather than as a composite measure such as 'suicidal acts' or 'suicidal behaviour'. The failure of studies to differentiate between different aspects of suicidal behaviour in reporting the outcome of interventions makes direct comparison between the impact of interventions on these diverse behaviours difficult. Looking only at those studies which focussed on a single outcome, reliably established positive outcomes were as follows:

G.33 The relatively low likelihood of success reported for studies evaluating outcomes specifically for self-harm is mirrored in the outcome of a regression analysis we carried out using all 200 studies. This analysis evaluated the relative contribution of key features of the studies to the overall outcome of whether or not a study reported a reduction in the behaviour evaluated following the intervention. Outcomes, for the purposes of this analysis, were judged to be positive if reductions in the behaviour were confirmed either by reference to statistical analysis or purely on the basis of the narrative report of study authors. This allowed us to include both quantitative and qualitative studies in the regression analysis. The characteristics of the studies entered into the initial regression equation were chosen to reflect issues of importance both to research and to clinical practice. Specifically, we took into account major features of study design (whether a study was quantitative or qualitative and whether it followed an RCT or other design, with the latter coded as zero for qualitative studies); the mode of intervention (pharmaceutical; psychotherapeutic/psychosocial; service-delivery); the population (general population vs other, psychiatric population vs non-psychiatric population) and setting (community vs other setting) in which the intervention was evaluated and the specific focus of the study (whether the reported outcomes referred to completed suicide, attempted suicide, self-harm, suicidal ideation or a combination of suicidal behaviours, the latter implied by terms such as 'suicidal acts' or 'suicidal behaviours').

G.34 The regression analysis was a stepwise logistic regression, with all dependent and independent variables coded as zero or one. Table G.9 overleaf sets out the percentages of studies falling into a category included in the regression equation. The Beta coefficients are a measure of the relative importance of each variable, and the significance levels indicate whether their contribution to outcomes was statistically significant or not. One study was excluded from the analysis as data for key variables were missing. To avoid further loss of power, the variable for setting (community versus other setting) was coded as 'community' in studies for which settings varied across time, provided the study either started or ended in the community.

Table G.9 Regression analysis to predict overall outcomes

G.35 In total just under one third of studies (26%) reported positive outcomes for at least one measure of suicidal behaviour or suicidal ideation. The outcomes in Table G9 provide a clear message regarding the characteristics of study design and focus which contribute independently to the likelihood of positive outcomes. Quantitative study designs (the majority of which were of substantially better quality than the available qualitative material) were more likely to result in positive outcomes for the evaluation of an intervention. Evaluations of interventions targeted at psychiatric populations were also significantly more likely to result in positive outcomes and studies focussed on self-harm, as opposed to other aspects of suicidal behaviour, were significantly less likely to result in positive outcomes.

G.36 Other associated factors not entered into the equation (for example the fact that quantitative study designs were more likely to use scale-based outcome measures than qualitative study designs) may have an impact here. Nevertheless, the regression analysis is informative. One key point to note is that, controlling for other factors, neither the broad mode of intervention nor the settings in which interventions were evaluated contributed to the final equation. The important issues were the population in which the intervention was evaluated, the form of behaviour which the intervention was intended to address and the use of a quantitative rather than a qualitative study design (in this context a quantitative design is likely to be a proxy for the overall quality of study design). The outcome of the regression equation therefore supports recommendations that the quality of future research is improved and that research urgently addresses the need to find effective interventions for people who self-harm and for people who are experiencing suicidal behaviour or suicidal ideation but do not have mental health problems.

The economic costs of intervention for suicidal behaviour and suicidal ideation

G.37 One of the objectives of the current review was to assess the available information on the cost and/or cost effectiveness of interventions. Unfortunately, only three of the studies included in the review addressed or refer to this issue (Vandersande et al 1997, Tyrer et al 2004, Duggan et al 2003). Since the efficacy of any given intervention clearly remains to be confirmed, the lack of focus on cost effectiveness is not surprising. The cost effectiveness of an intervention is inherently tied to the extent to which it produces positive outcomes. Since the literature is still at the stage of attempting to establish which interventions can be regarded as effective in any unambiguous way, it may be regarded as premature to begin evaluating cost effectiveness.

G.38 Further cost effectiveness studies are likely to follow as researchers begin to feel more confident regarding the clinical benefits of particular interventions. To draw on the limited cost-effectiveness evaluations currently available, a mathematical modelling study (Duggan et al 2003) funded by Novartis Pharmaceuticals reaches the conclusion that if we assume clozapine to be cost-neutral, the potential saving per life saved of prescribing clozapine for treatment resistant schizophrenia would be £5108. Assuming that clozapine achieves a 10% reduction in annual support costs, the net saving would be £8.7 million per year, with additional savings potentially deriving from an estimated reduction in the number of acute hospital beds used each year of 167. The authors conclude on this basis that clozapine is cost effective. Both the assumption of a cost-neutral scenario and the assumptions of percentage efficacy are, however, rather optimistic, notably the latter assumption, given the somewhat equivocal outcomes for clozapine demonstrated by the studies included in the current review.

G.39 A further study, relying on actual outcomes rather than mathematical modelling (Tyrer et al 2004, POPMACT study) presents data regarding the cost-effectiveness of Manual Assisted Cognitive Behaviour Therapy ( MACT) in the treatment of recurrent self-harm. It is worth noting that the study showed no significant difference between those repeating self-harm in the MACT group than in the treatment-as-usual ( TAU) group, so the practical value of estimates of cost-effectiveness for this approach is debateable. However, the authors conclude that the treatment was 10% cheaper than TAU. The implications for clinical decision-making are, however, more complex than this simple message suggests. The study distinguished between patients with and without borderline personality disorder. Whilst the authors report that the mean cost of MACT vs TAU over 1 year did not differ between BPD and non- BPD patients ( BPD cost of MACT £14,524 per person, cost of TAU £15,665/non- BPDMACT £12,618, TAU £13,331), the mean cost of treating BPD with either (£15,081) was significantly greater than treating non- BPD patients (£12,985, p<0.004). No differences in cost between the treatment of BPD and other personality disorders were noted. The separation of the available cost data by disorder in the only pertinent tables in the report also means that the cost differences cited are not available to the reader as absolute figures for further consideration.

G.40 The final study referring to issues of cost effectiveness (Vandersande 1997ps) provides only very limited speculation on the likely cost effectiveness of an intervention. The study itself relates to an intensive in-patient and community intervention following attempted suicide, drawing comparisons with routine care. In respect of clinical efficacy, the authors conclude that despite intensive intervention suicide attempts were not reduced in comparison with routine care and the number of days in in-patient psychiatric care was also not reduced. The issue of cost is only raised in respect of the increased contact with services identified for the experimental group The authors simply note that the cost implications of this would be slight, but provide no data to support this assumption. No specific evaluation of cost is carried out and since the intervention was not effective it is unclear that any further exploration of costs would be justified at this stage.

« Previous | Contents | Next »