Abstract
Keywords
Deliberate self-poisoning is one of the most common reasons for hospital admission in the UK. It accounts for 170 000 hospital attendances per annum [1]. Patients presenting to hospital following deliberate self-poisoning are at high risk of eventual suicide [2, 3] and one hundred times more likely to commit suicide than the general population [4].
There is no accepted or agreed treatment for deliberate self-harm, but psychological approaches involving problem solving have shown promising results [5]. Since over 70% of self-poisoning patients cite an interpersonal problem as the main precipitant of the episode, there is a strong rationale for using interpersonal therapies in this group. We have shown that brief psychodynamic-interpersonal therapy (PIT) in comparison with usual care results in a significant reduction of suicidal ideation and repetition of self-harm, 6 months post-treatment [6]. In this paper we explore factors that predict a good response to treatment.
Method
The research design, described elsewhere [6], involved a randomised controlled trial of psychotherapy versus usual care in patients presenting to an emergency department, in Manchester, England, following deliberate self-poisoning. Inclusion criteria were: age between 18 and 65 years; able to read and write English; living in the hospital's catchment area; registered with a general practitioner; and not requiring inpatient psychiatric treatment. Patients were assigned to the psychotherapy intervention or usual care in blocks (of 12 consecutive patients) using randomization lists provided by a statistician. The groups were stratified in terms of items of previous self-harm. Assessments were conducted at baseline, following the treatment phase (4 weeks), and at 6 months follow-up.
The intervention
Patients were offered four sessions of psychodynamic interpersonal (PI) therapy. This model, developed by Hobson [7], uses the patient– therapist relationship as a tool for resolving interpersonal issues. It has been found to have effects equivalent to cognitive behavioural therapy for the treatment of depression [8, 9]. The therapy, adapted for patients who have self-harmed, was given by one of three nurse therapists (SS, CT, GB) in the patient's home.
Hobson's model emphasizes a strong working alliance; key aspects include picking up cues, staying with feelings, and working in the ‘here and now’. The therapist adopts a negotiating style and is open to correction. The intention is to engage the patient in a conversation, or a ‘shared feeling language’ that moves gradually to deeper and deeper levels. Metaphor is used where ever possible to link images to underlying feelings. Statements are preferred to questions and are couched in the form of hypotheses. These are tentative but informed guesses that the therapist makes about the client's inner world. Further details of the model have been described by Guthrie [10]. Although many of these components are non-specific, together they form a definable therapy that is relatively easy to learn and understandable to patients.
In working with patients who have self-harmed, the first task is to explore the circumstances that precipitated the episode. Emphasis is placed on exploring feelings, and bringing these into the here and now. Problems, which have precipitated the self-harm, are explored and a rationale linking feelings, problems and relationships is developed.
Shared understanding is important in the process and is formed through therapist and patient clarifying what the latter is experiencing and feeling. In order to do this a ‘language of mutuality’ is developed; the therapist uses terms such as ‘I’ and ‘we’. The therapist is more active than in psychoanalytic therapy but many of the interventions are similar, especially metaphor and interpretation. Interpretations are construed as tentative hypotheses. The aim is to produce a meaningful dialogue in which the therapist's suggestions can be worked with but modified and owned by the patient.
Even in this brief work, issues that arise between therapist and patient are actively explored and linked to important relationships in the patient's life. Gaining understanding of, and assimilating problems are key aims, and include integration of past and present experience.
In working with people who have self-harmed, certain aspects are particularly useful. Picking up cues, staying with feelings and working in the here and now enable the therapist to carry out an in-depth assessment of suicidality. As some three-quarters of patients cite an interpersonal problem as the main precipitant, it is usually straightforward to identify an interpersonal problem that can be explored. Attention is paid to practical solutions as well as interpersonal change. Meaningful links between the patient–therapist relationship and relationships outside of therapy can also be made.
At the end of therapy, links can be made to other services (e.g. if appropriate, psychotherapy). One of the most useful functions of the therapy, in addition to addressing interpersonal issues and psychological distress, is to engage in treatment patients who would otherwise refuse further help or not attend hospital appointments. Most patients, however, are not referred on to mainstream general psychiatric services.
Usual care
Patients who were randomized to ‘treatment as usual’ received basic care, in most cases this was advice to consult their own general practitioner.
Main outcome
The main outcome measure, the Beck Scale for Suicidal Ideation (SSI) [11], was used as a proxy for suicide, since strength of suicidal ideation is an important predictor of completed suicide [12]. Depressive symptoms as measured by the Beck Depression Inventory (BDI) [13]; patients’ satisfaction with treatment and repetition of deliberate selfharm were considered secondary measures. Other variables measured at baseline but not used for outcome, were the Beck Hopelessness Scale (BHS) [12], the Beck Anxiety Inventory (BAI) [14] and the short form Inventory of Interpersonal Problems (IIP-32) [15].
Patients randomized to psychotherapy had a significantly greater reduction in suicidal ideation (p = 0.005) at 6 months post-treatment, than those who received usual care, they were more satisfied with their treatment, had a significant reduction in repetition of deliberate selfharm 6 months post-treatment (proportion repeating 8.6% vs 27.9%, p = 0.009, Fisher's exact test, 95% confidence interval for difference in proportion 8.6% to 30%).
Data analysis
The main analysis in this paper concerns prediction of outcome following psychotherapy; data are also presented for all participants. The outcome of interest was the suicidality of patients at 6 months; univariate analysis was performed to determine the variables associated with scores on the Beck Scale for Suicidal Ideation. Those variables that were significant, or nearly so (p < 0.2), were entered into a forward stepwise multiple regression with suicidal ideation at 6 months as the dependent variable.
Results
Characteristics of the patients at baseline
Of the 119 patients who entered the trial, 66 (55.5%) were female, and the mean age was 31.2 (SD = 11.5). Thirty-three patients were married or cohabiting (27.7%). Only 19 (16.0%) were in paid employment. Paracetamol was the drug most frequently chosen for selfpoisoning (36.1%). Psychological assessments were completed in 89 (74.8%) patients at the end of treatment, and in 95 (79.8%) at follow-up.
Predictor variables
The potential predictor variables are shown in Table 1. Sixty-three per cent of patients had high scores (greater than 25) on the Beck Depression Inventory and over half had a previous history of self-harm. The two treatment groups were similar on all baseline variables with the exception of marital status.
Baseline variables for psychotherapy and usual treatment groups
Univariate analysis
Eight variables were associated (p < 0.2) with SSI scores at followup (Table 2 shows the categorical variables) for the group as a whole, previous history of psychiatric disorder, marital status, previous history of deliberate self-harm, suicidal ideation (SSI) (p < 0.001), depression (BDI) (p < 0.001), anxiety (BAI) (p < 0.001), hopelessness (BHS) (p < 0.05) and interpersonal problems (IIP-32) (p < 0.01). In the psychotherapy group, seven variables were associated (p < 0.2) with SSI scores at follow-up: previous history of psychiatric disorder, alcohol taken during self-poisoning episode, previous history of deliberate selfharm, suicidal ideation (SSI) (p < 0.01), depression (BDI) (p < 0.01), hopelessness (BHS) (p < 0.2) and interpersonal problems (IIP-32) (p < 0.05).
Severity of suicidal ideation at 6 months according to baseline categorical variables (only variables that were associated with outcome (p < 0.2) are shown)
A prior history of self-harm and of psychiatric disorder were associated with more severe scores on the SSI at follow-up, both for the group as a whole and for the psychotherapy patients (Table 2).
Multivariate analyses
For the group as a whole, three variables (after controlling for treatment group) were selected into the multiple regression equation explaining 39% of the outcome (Table 3). Severity of suicidal ideation was the first variable, followed by anxiety and previous history of selfharm.
Stepwise multiple regression to predict suicidal ideation at 6 months for whole trial, and psychotherapy group
For the psychotherapy group, two variables were selected, explaining 27% of the variance: severity of depression and previous history of self-harm (Table 3).
Discussion
We chose suicidal ideation at follow-up as the main outcome measure. Although repetition of self-harm is an important outcome, only a minority of patients repeat self-harm in a 12-month period [15]. A focus on repetition alone means that other problems may be missed.
Psychotherapy appeared to be more effective for patients with lower depression scores and no prior history of selfharm, findings in accord with other trials in deliberate self-harm [16, 17], one small study using behaviour therapy was more effective for patients ‘initially reporting more severe depressive symptoms’ [18]. However, as the mean BDI score for patients entering the intervention was only 9, they are unlikely to be similar to our treatment group in which the mean BDI was three times that level.
Age or gender were not related to outcome. This is similar to some studies [16, 19], but other investigators [20] have reported that women have a better outcome than men, in terms of social functioning.
It is important to note, that in most studies repetition as self-harm is the main predictor variable. Differences in setting and clinic groups suggest that only tentative comparisons can be drawn between samples.
It is not surprising that baseline suicidal ideation was a predictor of suicidal ideation at 6 months for patients as a whole. Prior history of self-harm is a predictor of poor outcome regardless of intervention. Prior history of psychiatric disorder was associated with a poor outcome in the univariate analyses but was not included in either regression model. It is likely to be associated with other variables that were selected. Severity of interpersonal problems also correlated with suicidal ideation at 6 months but was not selected into either regression model.
It is important to remember that both regression models (for psychotherapy and whole groups) accounted for less than 40% of the variance, which suggests that many other factors not measured influenced outcome.
Our psychotherapy programme was very brief. Other investigators have shown that PIT has a linear dose– response in treating depression [21] (i.e. the greater the number of sessions, the greater the reduction in depressive symptoms). They also found that 16 sessions of PIT may have particular benefits for those with severe depression. This suggests that self-poisoning patients with more severe depressive symptoms may need more extended treatment.