The Scottish Government

« Previous | Contents | Next »

Listen

ANNEX F: OVERVIEW OF PRIMARY STUDIES RETRIEVED

Profile of Primary Studies Included

F.1 All but two of the 200 primary studies included in this review were identified by the review team through electronic searching. The exceptions (Thrive Initiative 2006; Gerber 2003) were unpublished evaluation reports provided by NIST. The majority of included material (54%) is of recent date (dated 2000-2006), with 37% of the material dating from the 1990s and only 8% of included material produced in the 1980s or earlier. This is in part due to the nature of electronic searching (although the majority of databases searched include material at least from the 1980s if not earlier) but it also reflects a trend for a rapid expansion in the literature addressing suicidal behaviour and suicidal ideation. In terms of the type of intervention focussed on in the literature, publication dates suggest a slight but significant shift over time (? ²=12.9 p<0.04) with an increase in the proportion of research focused on psychotherapeutic interventions and service delivery and a decrease in the proportion of studies focussed on pharmaceutical and other interventions.

F.2 The intervention literature is extremely diverse in respect of the approach taken to preventing suicidal behaviour and ideation. Within the 200 studies included, we identified over 150 different specific interventions which had been evaluated. This level of diversity, accompanied by the more restricted but nevertheless diverse range of settings, populations and age groups in which the interventions have been evaluated precludes any meaningful meta-analysis in the context of the current report. With the benefit of additional information or, preferably, individual level data from study authors, limited meta-analysis to address specific issues may be a possibility in the future. The broad categories of intervention addressed and their contribution to the total number of studies included are set out in Table F1 below.

Table F.1 Approaches to Intervention identified in the Literature

F.3 Although there has recently been a slight decrease in the proportion of research focussed on pharmaceutical intervention, this area clearly remains the driving force behind research in the prevention of suicidal behaviour and ideation. As a contributor to the overall body of knowledge it comes close to being rivalled only by the very eclectic collection of 'other' interventions which have been evaluated and these are too diverse to be regarded as directing the literature in any meaningful way. Note here that given our very broad inclusion criteria, there is little reason to assume that the studies we have included present a particularly biased view of the empirical intervention literature, with the possible exception of the foreign-language literature, which for pragmatic reasons, we specifically excluded.

F.4 Although foreign language literature was excluded, the range of countries from which the included studies originated is broad. Excluding multi-national studies, populations from 21 countries provided studies for the review. Note here, that where study authors reported on a population other than their native population, we coded study origin in respect of the population for which an intervention was evaluated, not the country of origin of the author of the research. Table F.2 below provides an overview of the origin of included studies compressed into categories.

Table F.2 Country of Origin of Included Studies

F.5 As with the majority of public health research and, indeed, research more generally, the population to which the majority of evidence applies is the United States (only seven of the studies included within this combined category derived from Canada). In comparison with other public health literatures, including comparable literature on other-directed violence, the proportion of the intervention research carried out within the UK is quite high, slightly exceeding that of research carried out in other European countries. This is likely to be due, in part at least, to the longstanding work of specialist research centres such as the University of Oxford Centre for Suicide Research. Of the 38 studies carried out in the UK, 8 related directly to the Scottish population. However, of these, four studies (Davidson et al 2004, Evans et al 1999, Tyrer et al 2003 and Tyrer et al 2004) reported on different aspects of the same multicentre trial (the POPMACT study). Only one of the five centres taking part in the study was sited in Scotland. Further details of outcomes from these linked studies and from the four additional and independent Scottish studies (Cunningham-Owens et al 2001, Eagles et al 2003, Gerber 2003 and Thrive Initaitive2006) are provided in the main text of the report.

F.6 In terms of the specific populations addressed by the included studies, the majority of studies, as with the majority of prior systematic reviews, focussed on psychiatric populations (46% of all included studies). This distribution is likely to be out of kilter with the true distribution of suicidal behaviour and self-harm, although evidence addressing the incidence of suicidal behaviour (other than completed suicide) in the general population, particularly in the UK, is vanishingly small and an accurate estimate of prevalence remains to be established. Within the group of studies addressing suicidal behaviour and ideation in people with a psychiatric disorder, the literature showed a clear bias towards specific types of disorder. The proportion of studies on psychiatric patients falling into each category was as follows:

F.7 The strong bias towards an evaluation of interventions for suicidal behaviour and suicidal ideation in the context of depression is unsurprising. Both this focus and the focus on schizophrenia are broadly in line with lifetime risk of suicide in these disorders in comparison to estimates for the general population ²², although the emphasis on depression in particular remains exaggerated in purely numerical terms. The focus on personality disorder may reflect a confounding of definitions. Whilst a number of studies (e.g. Bronisch 1996, Duberstein 1997) have cited the apparently close association between personality disorder and likelihood of suicide, estimates of lifetime prevalence of suicide in this disorder commonly fail to take into account that suicidal behaviour remains one of the defining criteria for a diagnosis of personality disorder. Without controlling for this definitional circularity it is difficult to establish the true association between this disorder and suicidal behaviour and hence to establish what the appropriate balance of research effort should be.

F.8 Outside of the psychiatric population, the main population focus of research (30% of included studies) is, again unsurprisingly, people presenting with suicidal behaviour and, to a lesser extent, ideation, or people otherwise seen as at high risk of suicide or suicidal behaviour. Amongst those studies specifying a particular group within the general population as the focus of intervention, the single dominant population group (12 of 14 such studies, 86%) of interest is adolescents. This is largely justified by population rates of suicide, which consistently indicate the relatively high risk of suicide in adolescent populations. In contrast, the impact of suicide on children and older people is not reflected in the amount of available research on interventions for these populations. Other specific populations on which a small number of studies have focussed in evaluating interventions (7% of included studies) are military personnel, the prison population, people with physical disorders and people who misuse substances.

F.9 Taken as a whole, the age range of participants in all included studies (where such information was provided) was between 6 and 94. In terms of the age ranges specified by the Research Advisory Group, the distribution of those studies providing age details and focussing on participants from particular age groups is as follows:

F.10 A substantial proportion of studies (29%) failed to provide any details of the age of their participants, the remainder recruited participants from across one or more or the age ranges specified above. The failure of primary studies to provide key demographic information is not an uncommon finding of systematic reviews. In the current context, reporting of the gender of participants was comparatively good ('only' 13% of the studies failed to provide such information), but information relating specifically to either women or men remains scarce, as the majority (73%) of studies providing information on the gender of their participants included mixed groups, with no separate sub-analyses of outcomes for males and females.

F.11 In line with the emphasis on borderline personality disorder, studies focussing solely on interventions with women (11%) were more common than studies focussed solely on men (5%). As we found also in our previous review of other-directed violence, participant ethnicity was particularly poorly reported by the included studies. In total, 68% of the 200 studies failed to provide information regarding the ethnicity of their participants. Of those studies which did provide this information, 84% included mixed ethnicities, with no separate sub-analysis by ethnic group. Out of the remaining 10 studies, six addressed interventions with participants drawn from minority ethnic groups only, the remaining studies focussing purely on white participants.

F.12 With regard to other key characteristics, nearly half (46%) of the studies failed to identify whether prior suicidal behaviour or ideation had been identified in their participants (that is, whether the behaviour used as an outcome measure represented a first time incident or a repetition of behaviour previously engaged in). Of the remaining studies, only two reported that the behaviour represented a first-time incident. For an additional seven studies the only behaviour known to have occurred previously was suicidal ideation. This means that 49% of the 200 included studies provide evidence for interventions evaluated in the context of repeated rather than first time suicidal behaviour.

F.13 Notwithstanding the high proportion of studies focussed on participants with a history of suicidal behaviour, the failure of nearly half of the available studies to consider and/or report the previous history of their participants precludes accurate analysis of differences between 'repeaters' and non-repeaters'. As a proxy, we will compare outcomes for studies which specifically state that their participants had engaged in actual acts of suicidal or self-harming behaviour and studies which did not record this or which failed to specify whether or not participants had engaged in such behaviour prior to entry into the study. Despite the poor reporting of this key characteristic, it seems there is a clear and, given their known risk status, an appropriate focus in the literature on people identified as having engaged in prior suicidal behaviour.

F.14 The Research Advisory Group expressed a specific interest also in studies addressing intervention for people known to misuse substances. Although only one study (Ahrens et al 1993) explicitly set out to address intervention for people identified as 'substance abusers', 16% of studies noted that one or more participants were 'substance abusers' and 12% of studies identified participants as having a diagnosis of substance abuse. A number of points can be made here. Firstly, as with other participant details, the key feature likely to interact with outcomes for suicidal behaviour has been left out of the equation by the majority of studies. In total, 73% of the included studies failed to identify whether or not participants were known to be abusing (or even using) alcohol or illicit substances during the course of their study and 79% of studies failed to identify whether or not one or more participants had a recorded diagnosis of substance abuse. Again, this is not an uncommon finding in systematic reviews; however, it is particularly unfortunate in the current context, where there is a known association between the outcome of interest and the use of alcohol and other substances.

F.15 In terms of study design, the review, as intended, identified a broad range of quantitative, quasi-experimental and qualitative studies. Dividing these often highly individualistic designs into broad categories, the literature, as represented by the studies meeting our inclusion criteria, again demonstrates a clear bias in approach. The proportion of studies falling into each broad category is outlined below:

F.16 Surprisingly, the high proportion of randomised controlled trials in the literature is driven not by the preponderance of pharmaceutical trials, but by the significantly higher proportion of psychotherapeutic/psychosocial studies following an RCT design (? ²=24.7 p<0.001). Whilst 44% of pharmaceutical trials adopted an RCT methodology, the majority of psychotherapeutic/psychosocial interventions did so (64%). Since around 23% of both service delivery and 'other' intervention studies also followed this methodology, it can reasonably be said that this literature displays a comparatively sophisticated methodological approach. Whilst the universal applicability of the RCT approach to evaluating all forms of intervention has been challenged (e.g. Weightman et al 2005) it currently remains the 'gold standard' approach in evidence-based health care. The widespread use of this methodology in evaluating interventions for suicide and self-harm is in notable contrast to other relevant public health literatures (cf. Leitner et al 2006).

F.17 Whilst qualitative designs are not wholly lacking, the number of detailed and well-conducted in-depth studies of the 'lived experience' of intervention for suicidal behaviour and ideation is disappointing. The design and implementation of the qualitative studies available is also generally of a rather lower standard than that of the quantitative research. The majority of the qualitative studies included (63%) are case studies and the vast majority of these are brief narrative accounts of a case intervention, with little structure and few attempts to validate the conclusions reached using any accepted principles of qualitative investigation. Of the remainder, only three studies follow an approach which could be considered to follow pre-defined principles of qualitative methodology (content analysis, non-participant observation and psychological autopsy). All other studies falling within the qualitative category simply report outcomes based on interview or survey data, largely in the absence of any structured approach to the investigation of stated hypotheses.

F.18 A particular strength of study design in this literature is the relatively large sample sizes achieved. The median sample size for all included studies is 140, with 55% of studies reporting an initial sample size of 100+. In addition, the majority (65%) of studies used prospective follow-up, reported statistical analyses where appropriate (70%), followed an intention-to-treat analysis (56%), used a small number of pertinent outcome measures (59% used three or fewer outcome measures) and, again where appropriate, reported baselines for all main outcome measures (56%). These are all characteristics which provide some confidence in the likely reliability and validity of the outcomes reported by the intervention studies.

F.19 Some differences in 'quality markers' were noted between studies with a focus on different types of intervention. Intention-to-treat analyses were more likely to be carried out in pharmaceutical studies than in studies focussed on all other modes of intervention (? ²=37.4 p<0.001) and psychotherapeutic/psychosocial studies were likely to have a higher number of outcome measures than pharmaceutical studies (a mean of 5.56 versus 2.61 in pharmaceutical studies t= -3.05 p<0.004). These differences, although worth noting, are not of sufficient weight to warrant blanket conclusions regarding the relative quality of studies addressing different modes of intervention. Overall, the quality of studies evaluating distinct modes of intervention is broadly comparable. This again differentiates the current research literature from other public health literatures, where the quality of study design varies more markedly with the interventions evaluated (cf. Leitner et al 2006).

F.20 In addition to the intrinsic features of study design set out above, the value of the research literature to clinical decision-making depends also on the settings in which interventions have been evaluated. In contrast to details given regarding participant characteristics, the majority of studies provided a clear description of the setting in which interventions took place. Only 8% of studies failed to provide any details regarding setting, with a further 2% providing details of the setting at the start of the study but failing to clarify whether or not all participants remained in the same setting during follow-up. Twenty-two studies (11%) involved participants drawn from a mixed range of settings. Taken together, this leaves 78% of the studies providing clear information regarding the settings in which interventions have been evaluated. Of these studies the vast majority (88%) evaluated outcomes with baseline and follow-up in a single setting. The distribution of settings for these single-site studies is as follows:

F.21 Of the 15 studies with distinct start and end settings, 12 began in an in-patient psychiatric setting with follow-up into the community, two began in A&E with follow-up into the community and one began in a community setting with follow-up into in-patient psychiatric care. In line with the focus on specific population groups outlined earlier, 'other' settings included prison, military bases and outpatient units involved with physical rather than mental health care. Whilst the balance of the above distribution is not wholly disproportionate to the balance of settings in which clients find themselves, some aspects of the above profile are a matter for concern. In particular, the lack of studies taking place in A&E, which provides a very significant point of first contact for a substantial number of people with suicidal behaviour and ideation. Similarly, given the comparatively high rates of suicidal behaviours in prisons, it is of considerable importance that further intervention studies focussed specifically on this setting are carried out.

F.22 Given the focus of the review on studies which addressed suicidal behaviour or suicidal ideation as an explicit outcome measure, it is not surprising that in the majority of the included studies (79%) the main focus was on suicidal behaviour or ideation. The main focus in the remainder of the studies was on interventions targeted at depression (16%), all-cause mortality (1%) or on outcome measures of no direct relevance to suicidal behaviour but with specific reference to the benefits or otherwise of the intervention for also reducing suicidal behaviour or ideation (3%). Note that only two of the studies falling into the latter category evaluated pharmaceutical interventions.

F.23 With regard to the particular form of suicidal behaviour addressed by the studies, nearly half (47%) focussed either solely on suicidal ideation or used this as a subsidiary outcome measure. Completed suicide was used as an outcome measure by 34% of studies, with 37% focussing on attempted suicide (variously defined) and 22% focussing on self-harm. Around 40% of studies included more than one of the above methods, primarily combining suicidal ideation as a subsidiary outcome with a main focus on suicidal acts, in particular attempted suicide or self-harm. Despite the fact that the majority of studies addressed suicidal behaviour or ideation as their main focus, comparatively few studies focussed solely on these issues, with 62% including also other outcomes in particular general mental health and social functioning.

F.24 Studies used a variety of methods to establish outcomes for suicidal behaviour and ideation. Official statistics were the most common source of information in respect of completed suicide (61% of studies evaluating outcomes for suicide used official statistics for suicide as the specific outcome measure). A more diverse range of sources was used to establish outcomes for self-harm, the most prominent approaches being via hospital or other formal records (24%) or narrative and scale-based self-report by participants (20% in each case). 'Attempted suicide' was the outcome measure least clearly defined by study authors and in line with this, a fifth of studies (20%) also failed to state how the outcome had been identified. This is a particular problem in the literature, since behaviour defined by a study author as 'attempted suicide' ranges across the full spectrum from minor self-harm to an act with near-fatal consequences. Parity of definitions for this and other aspects of suicidal behaviour and ideation or, at least, clear descriptions of the specific behaviour addressed by a research study would have the potential to significantly improve the value of future research for intervention practice.

F.25 A high proportion of studies evaluating outcomes for self-harm (54%) addressed only attempted self-harm. A further third (27%) addressed both actual and attempted self-harm, with the remaining 18% addressing actual self-harm only. As with attempted suicide, more precise descriptions of the behaviour referred to were scarce, with 41% of studies failing to provide any further definition and 34% using the eclectic term 'any method' of self-harm as their identified outcome measure, without further sub-analysis by means. The only methods of self-harm for which further detail was generally provided were self-mutilation (18% of studies using self-harm as an outcome) and self-poisoning (7% of studies evaluating outcomes on the basis of incidents self-harm).

F.26 Contrasting what is known about the incidence of different forms of suicidal behaviour with patterns observed in the intervention studies included here, the proportion of research addressing rarer forms of suicidal behaviour (suicide, attempted suicide) is out of balance with the incidence of observed behaviour. Whilst it is clearly the case that completed suicide, in particular, is a more severe outcome, it seems to be the case that self-harm, as a common, or possibly very common behaviour, in the general population and in particular in young populations (c.f. Fox & Hawton 2004) is under-researched by comparison. From the limited description of behaviours provided in the included studies addressing self-harm it seems also that self-cutting is under-represented as a specific focus of attention. The use of multiple methods of self-harm, together or sequentially over time is a form of self-harm barely referred to in the context of intervention studies. It may be possible to explore this issue further by contacting study authors for further more detailed information regarding the forms of self-harm carried out by participants.

« Previous | Contents | Next »