Article Text

Download PDFPDF

Randomised controlled trial of therapeutic assessment versus usual assessment in adolescents with self-harm: 2-year follow-up
  1. Dennis Ougrin1,
  2. Isabel Boege2,
  3. Daniel Stahl3,
  4. Reetoodwaj Banarsee4,
  5. Eric Taylor1
  1. 1Department of Child and Adolescent Psychiatry, King's College London, Institute of Psychiatry, London, UK
  2. 2ZfP Suedwuerttemberg, KJPP, Ravensburg, Germany
  3. 3Department of Biostatistics, King's College London, Institute of Psychiatry, London, UK
  4. 4Department of Primary Care and Public Health, Faculty of Medicine, Imperial College London, London, UK
  1. Correspondence to Dr Dennis Ougrin, Child and Adolescent Psychiatrist, King's College London, Institute of Psychiatry,  DeCrespigny Park, Box 85, London SE5 8AF, UK; dennis.ougrin{at}


Background An earlier randomised controlled trial demonstrated improved treatment engagement in adolescents who received Therapeutic Assessment (TA) versus Assessment As Usual (AAU), following an emergency presentation with self-harm.

Objectives To determine 2-year outcomes for the same adolescents focusing on frequency of Accident and Emergency (A&E) self-harm presentations and treatment engagement.

Method Patients in the TA groups (n=35) and the AAU group (n=34) were followed up 2 years after the initial assessment. Their primary and secondary care electronic records were analysed.

Results There was no significant difference in the frequency of self-harm resulting in A&E presentations between the two groups (OR 0.69, 95% CI 0.23 to 2.13, p=0.53). Treatment engagement remained higher in the TA group than the AAU group.

Conclusions TA is not associated with a lower frequency of A&E self-harm presentations. The effect of TA on engagement is maintained 2 years after the initial assessment. Interventions to reduce self-harm in adolescents are needed.

Trial registration ISRCTN 81605131,

  • Accident & Emergency
  • Child Psychiatry

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

What is already known on this topic

  • 10% of adolescents will have self-harmed by the time they finish secondary school and 10% of the adolescents who engage in self-harm will repeat self-harm within a year.

  • Adolescents with self-harm have poor engagement with treatment.

  • Therapeutic Assessment and the Family Intervention for Suicide Prevention may lead to improved engagement with aftercare in the short term.

What this study adds

  • Therapeutic Assessment may lead to improved engagement with treatment over a period of 2 years.

  • Better engagement with treatment does not appear to lead to fewer emergency self-harm presentations.

Suicide is the second or the third leading cause of death in adolescents in most Western countries.1–3 Self-harm is the strongest predictor of eventual death by suicide in adolescence, increasing the risk up to tenfold.4 Poor engagement is an important problem in adolescents with self-harm5 and it is associated with poor outcomes.6 Trial of Therapeutic Assessment in London (TOTAL) compared the impact of Therapeutic Assessment (TA) and Assessment As Usual (AAU). TA is a brief therapeutic intervention based on cognitive analytic therapy7 delivered during the initial self-harm assessment. In TOTAL, 3 months after the initial assessment TA resulted in better treatment engagement than AAU but there was no difference between the two groups in the frequency of self-harm.8 This study looked at whether the two groups differed in the frequency of self-harm leading to Accident and Emergency (A&E) presentations over a two-year period as well as whether the improved engagement was maintained 2 years after the initial randomisation.


Study hypothesis

The main hypothesis of this study was that the number of adolescents with one or more A&E presentations with self-harm will be statistically significantly fewer in the TA than the AAU arm of TOTAL 2 years after initial randomisation.


The original trial recruited 70 adolescents aged 12–18 years who were not engaged with psychiatric services and who had self-harmed and been referred for a psychosocial assessment (table 1). Self-harm was defined as self-injury or self-poisoning irrespective of the underlying intent.9 Exclusion criteria were gross reality distortion (eg, owing to psychotic illness or intoxication), known history of moderate or severe intellectual disability, lack of fluent English, immediate risk of violence or suicide and the need for inpatient psychiatric admission. Participants received either TA or AAU at random. Randomisation occurred at the assessor level.8 The randomisation was stratified by centre, and two blocks (block lengths 22 and 4) were created using a permuted block design to ensure equal numbers of clinicians from each centre being allocated to either intervention or control groups. The adolescents who agreed to participate (n=70) did not differ significantly from those who disagreed (n=3) on any clinical or demographic characteristic.

Table 1

Participants’ baseline characteristics

The settings and locations where the data were collected

The referral for psychosocial assessment was made either following a screening at the emergency departments of four inner-London hospitals or following an urgent general practitioner's referral to the child and adolescent mental health services (CAMHS) in two London National Health Service Trusts. The referring practitioner and the emergency department staff were blind to the allocation of the adolescents to either TA or AAU.

Treatment interventions in the original study

Control group: assessment as usual

AAU included a standard psychosocial history and risk assessment and followed the recommendations set out in the National Institute for Health and Clinical Excellence (NICE) guidelines.9 The assessment letter was sent to the relevant community team, and a copy was sent to the family in accordance with the trusts’ policies. A random sample of 10 (29%) audiotaped assessments was assessed by two independent raters, and adequate adherence to the model was documented.10

Intervention group: therapeutic assessment

The major components of TA are as follows:

  1. Standard psychosocial history and risk assessment (approximately 1 h).

  2. A 10-min break to review the information gathered and to prepare for the rest of the session, followed by a 30-min intervention covering the next four steps.

  3. Joint construction of a diagram (based on the Cognitive Analytic Therapy paradigm) that consists of three elements: reciprocal roles, core pain and maladaptive procedures.11

  4. Identifying a target problem.

  5. Considering and enhancing motivation for change.

  6. Exploring potential ‘exits’ (ie, ways of breaking the vicious cycles identified).

  7. Describing the diagram and the exits in an ‘understanding letter’. In addition to the ‘understanding letter’, the family also received the usual assessment letter.

The assessment process was manualised, although assessing clinicians used clinical judgement when deciding on the best approach to ‘exits’. Family members were involved in all stages of TA whenever possible. Clinicians received five half-day training in TA accompanied by weekly homework and a video assessment before and after training with independent fidelity assessment.

Sample size

Sample size was not calculated for the primary outcome measure of this follow-up study. The sample size in the original study was calculated for engagement as the primary outcome measure.

Further assessments in this study

Data collection

Two years after the initial presentation, the participants’ health records were reviewed. Sources of the records included (a) electronic patient records of the CAMHS, (b) primary care records and (c) A&E records of the hospitals serving the area. No patient was excluded from the follow-up.

Data were collected on the number of appointments attended after the initial assessment, the number of missed appointments, the number of A&E presentations, hospital admissions and the total recorded number of self-harm episodes.


It was not possible to blind the clinicians to the intervention they were delivering. Participants were unaware as to what type of assessment they were receiving. The study statistician and the researchers who were conducting follow-up assessments were unaware of the participants’ allocation.

Primary outcome

Number of adolescents with one or more presentation to A&E with self-harm.

Secondary outcomes

(1) Attendance of treatment sessions in CAMHS, (2) episodes of self-harm not leading to A&E presentation, (3) proportion of the adolescents who continued to self-harm in year 1 and in year 2 and (4) total number of A&E presentations.

Clinical diagnosis

Clinical diagnosis was recorded using International Classification of Diseases, 10th edition criteria. Primary clinical diagnoses were collapsed to form four groups: (1) no diagnosis, (2) emotional disorders (primarily mood and anxiety disorders), (3) disruptive disorders (primarily conduct disorders and ADHD) and (4) other.

Statistical analysis

A random effects logistic regression was used to evaluate the occurrence of self-harm between intervention and control group. Patients were not sampled individually but as a group treated by a clinician. To adjust for possible clinicians’ effects, we included clinician as a random factor in the model that explicitly models the possible correlation between patients of the same clinician. ORs and 95% CIs are reported.

Random effects Poisson regressions were used to evaluate differences in the number of events within the observation period between intervention and control group.

We used robust SEs to adjust SEs for overdispersion (access of 0's observations) and for violations of the assumptions that the variance equals the mean in general. Again, to adjust for possible clinicians’ effects, clinician was included as a random factor. Results are reported as estimated incidence rate ratios (IRRs) with 95% CIs of comparing intervention to control group. The IRR is an estimate how many times greater the rate for events in the intervention group is compared to the control group.

Analyses were done using STATA V.11.1.12 The user-written STATA function gllamm was used for random effects Poisson regressions.

ITT analysis

Sixty-nine out of 70 recruited participants were analysed on the primary and secondary outcome measures. One participant (in the control arm) withdrew his consent for participation in the trial.


Primary outcome measure

Overall, the sample comprised 69 participants (one patient withdrew his consent to participate). Primary outcome data were available for 35 persons in the treatment group (28 female) and 34 in the control group (28 female); however, five adolescents moved out of the area during the follow-up period. In the intervention group, seven out of 35 patients presented to A&E with at least one episode of self-harm (three adolescents presented two, three and four times, respectively) while nine out of 34 patients presented with at least one episode of self-harm in the control group (one adolescent presented eight times). There was no statistically significant difference in the number of adolescents who presented to A&E with self-harm between the intervention and the control groups (OR 0.69 (95% CI 0.23 to 2.13), z= − 0.64, p=0.53). Predicted marginal probabilities to perform self-harm were 0.2 (95% CI 0.07 to 0.33) in the intervention group and 0.27 (95% CI 0.12 to 0.41) in the control group. There was no effect of the clinician on self-harm (ICC=0; 95% bootstrap CI 0 to 0.003).

Inter-rater reliability

Ten randomly selected clinical records were rated by the two trial researchers. Inter-rater reliability was assessed using the presence or absence of at least one episode of self-harm as the criterion. There was a good agreement between the raters, κ=0.8, 95% CI 0.43 to 1, p<0.02.

Secondary analyses

Engagement in treatment

A random effects Poisson regression with the number of psychiatric treatment sessions attended as the dependent variable, while time (year 1 or 2), treatment group and the interaction between time and treatment group as fixed factors as well as clinician as random factor did not reveal a significant interaction between group and time (p=0.27). Interaction was therefore removed from the model. The final model revealed that adolescents were more likely to engage whether they were in the TA during the first year of treatment (IRR 3.23 (95% CI 1.49 to 7.05), z=2.97, p=0.003). Engagement remained higher for TA participants in the second year (IRR 1.67 (95% CI 1.22 to 2.28), z=3.22, p=0.001).

Total number of recorded self-harm episodes

The total number of self-harm episodes was compared between the groups, after controlling for differences in the number of days open to the relevant service. A random effects Poisson regression with total number of self-harm episodes as the dependent variable and group as the fixed factor was used. The number of days participants were treated by CAMHS professionals differed between the groups (Intervention group: 303.2 days (SD 286.6), Control group: 243.65 days (SD 265.6)) and was included as a confounding variable in the regression model. Controlling for the ‘number of days treated’, the groups did not differ on the number of recorded self-harm episodes significantly (relavative risk (RR) 4.78, 95% CI 0.76 to 32.65, z=1.67, p=0.09).

Frequency of self-harm in year 1 and year 2

In the first year, 49 (71%) of the 68 adolescents did not have recorded self-harm episodes. The other 20 adolescents reported between one and 129 episodes (median=9.5; lower and upper quartiles: 1–33.5). In year 2, 55 (80%) did not have recorded self-harm episodes while the remaining 14 adolescents reported between one and 144 episodes (median=4.5 lower and upper quartiles: 1–12). A random effects Poisson regression with number of reported self-harm episodes per year as dependent variable, group as fixed factors and clinician and participant as random factors did not reveal significant differences between year 1 and 2 (IRR 0.56 (95% CI 0.19 to 1.66), z=−1.04, p=0.30).

Total number of A&E presentations (self-harm and other reasons)

There was no significant difference in the total number of A&E presentation (self-harm and other events) between intervention and control group (IRR 1.20 (95% CI 0.58 to 2.47), z=0.5, p=0.62, random effect: therapist).


Main findings

Two years after randomisation, there were no differences in the number of adolescents who presented to A&E departments with self-harm between the TA and the AAU arms. It may be that the study was not sufficiently powered to detect a difference in this outcome measure. The original study was powered only to detect a difference in engagement, and attrition may have further reduced the power to detect a clinically significant difference. If the difference detected (OR ∼ 0.7) were to hold true, a significantly larger sample size would have been required for this study to be adequately powered.

Nevertheless, the sample size is comparable to most other studies in the field, and therefore, this lack of difference appears important. It is therefore possible that the interventions offered to adolescents with self-harm in routine clinical practice are not effective in reducing self-harm. The authors of a recently published review concluded that there are no independently replicated studies demonstrating a significant reduction in self-harm in adolescents,13 thus highlighting the urgent need to develop effective interventions for adolescents presenting with self-harm. No brief interventions at the point of the initial assessment have been shown to reduce self-harm in adolescents to date. Adult literature points towards Dialectic Behaviour Therapy and other CBTs as potentially promising interventions but these require rigorous evaluation in adolescent RCTs.13

In this study, TA was associated with a statistically significant increase in treatment engagement of the adolescents presenting with self-harm over a period of 2 years. This study adds to the growing evidence that brief interventions for adolescents who present in crisis improve engagement. The findings indicate that the use of a therapeutically orientated initial assessment can greatly increase the chances of maintaining adolescents in follow-up therapy. TA might have lead to improved engagement by addressing the participants’ hopes and expectations from self-harm assessment through an unusual reminder (sending a TA understanding letter) or through improving adolescents’ experience of self-harm assessment. Alternatively, it is possible that spending more time with adolescents (approximately 40 min on average10) and clinicians’ enthusiasm may both have contributed to the increased engagement. In comparison with older similar studies investigating the impact of brief crisis interventions in adolescents presenting with self-harm, the improved engagement was greater yet the study was embedded into routine clinical practice and the intervention was delivered by front-line clinicians with no prior experience of research.

Spirito et al14 found that a treatment adherence enhancement intervention did not significantly increase the number of sessions attended (mean 7.7 sessions vs 6.4 sessions), although the findings reached statistical significance when adjusted for barriers to services (mean 8.4 vs 5.8). In a study by Rotheram-Borus et al15, the corresponding findings were 5.73 vs 4.67. The participants were also more likely to attend their first follow-up appointment in that study (95.4% vs 82.7%, p<0.02).15

Another example of a brief intervention associated with improved engagement with follow-up treatment (92% vs 76%, p<0.01 among youths receiving emergency evaluation for suicidality) is the Family Intervention for Suicide Prevention (FISP) based on Cognitive Behaviour Therapy.16 The FISP aims to use the emergency visit as a ‘window of opportunity’ to deliver a beneficial intervention and enhance motivation to attend follow-up treatment.17 TA and FISP point towards the importance of using the initial contact with self-harming adolescents therapeutically in order to enhance long-term engagement.

The finding of no significant reduction in the proportion of the adolescents who self-harm between year 1 and year 2 in the entire sample is of interest. It may be that the duration of the follow-up was not long enough to demonstrate a widely reported finding of a significant reduction in the frequency of self-harm in late adolescence and early adulthood.18 Nonetheless, the proportion of adolescents who self-harmed has reduced from 0.3 to 0.2 in the entire sample. There were no completed suicides in either arm, and no deaths due to other causes were recorded.


The design limitations of the small sample size and exclusion of clinically relevant subgroups all of which limit the power to detect differences in outcome and the generalisibilty of the findings were discussed above. In addition, we included adolescents with suicidal as well as non-suicidal self-harm in this study. Although many researchers in the UK and Europe consider self-harm to be a broad range of behaviours irrespective of the underlying intent, many American researchers study adolescents with suicidal and non-suicidal self-harm separately. It is indeed possible that suicidal adolescents might differ in some respects from non-suicidal self-harmers.19–21 Our sample size was too small to allow a more detailed analysis of subgroups. Our findings therefore may not be applicable to strictly defined suicidal or non-suicidal samples.

A further limitation of this study is not assessing several potential predictors of engagement including barriers to services, engagement with other (non-statutory) agencies and the family expectations regarding future therapy.

Main strengths of the study include pragmatic design and using an intervention with minimal training requirements10 delivered by front line clinicians.

Finally, it is possible that we did not record some episodes of self-harm due to five participants moving away from the area but this is unlikely to have changed the main conclusions of this study.


The authors thank the participants and their families for taking part in this study; the clinicians for their time and dedication in delivering the interventions; Derek Bolton and Robert Goodman for research advice, Peter Hindley, James Low and Gordana Milavic for clinical advice and supervision; Ozlem Eylem for database management; Nathan Parnell for editing this paper and Jo Fletcher for providing management advice and support. The authors would like to thank clinicians from South London and Maudsley NHS Trust, Royal Free Hampsted NHS Trust and West London NHS Trust for taking part in this trial. The authors would also like to thank the management of St Thomas’, Lewisham, King's College and Royal Free hospitals for facilitating patients’ recruitment.



  • Contributors DO, IB and ET designed the study. DO devised Therapeutic Assessment. RB and DS provided help with statistical analysis and interpretation of the results. All contributors reviewed and edited the manuscript. DO and IB are the guarantors.

  • Funding The study was funded from the following two sources: Psychiatry Research Fund (Institute of Psychiatry, King's College London) and Maudsley Charity (South London and Maudsley NHS Trust). The funders had no role in the study design in the collection, analysis and interpretation of data; in the writing of the report; and in the decision to submit the article for publication. The researchers were independent from funders.

  • Competing interests None.

  • Ethics approval This study was approved by the Camden and Islington Community Local Research Ethics Committee on 23 October 2007 (ref: 07/H0722/66).

  • Provenance and peer review Not commissioned; externally peer reviewed.

Linked Articles