Intervenciones conductuales y cognitivo‐conductuales para la conducta agresiva dirigida a terceros en personas con discapacidad intelectual
Resumen
Antecedentes
La conducta agresiva dirigida a terceros en personas con discapacidad intelectual es un problema importante que puede conllevar una mala calidad de vida, la exclusión social y el internamiento psiquiátrico. Se han desarrollado enfoques cognitivos y conductuales para controlar la conducta agresiva, pero la eficacia de estas intervenciones para reducirla y otros desenlaces no está clara. Esta es la tercera actualización de esta revisión en la que se añaden nueve estudios nuevos, para un total de 15 estudios.
Objetivos
Evaluar la eficacia de las intervenciones conductuales y cognitivo‐conductuales sobre la conducta agresiva dirigida a terceros en comparación con la atención habitual, los controles en lista de espera o ningún tratamiento en personas con discapacidad intelectual. También se evaluaron las intervenciones mejoradas en comparación con las no mejoradas.
Métodos de búsqueda
Se utilizaron los métodos exhaustivos estándares de búsqueda de Cochrane. La última fecha de búsqueda fue en marzo de 2022. Se corrigieron los términos de búsqueda para incluir el apoyo conductual positivo (ACP).
Criterios de selección
Se incluyeron ensayos aleatorizados y cuasialeatorizados de niños y adultos con discapacidad intelectual de cualquier duración, ámbito y cualquier comparador elegible.
Obtención y análisis de los datos
Se utilizaron los métodos estándar de Cochrane. Los desenlaces principales fueron el cambio en 1. la conducta agresiva, 2. la capacidad de controlar la ira, 3. el funcionamiento adaptativo y 4. los efectos adversos. Los desenlaces secundarios fueron los cambios en 5. el estado mental, 6. la medicación, 7. las necesidades asistenciales y 8. la calidad de vida, así como 9. la frecuencia de utilización de los servicios y 10. los datos de satisfacción de los usuarios. Se utilizó el método GRADE para evaluar la calidad de la evidencia de cada desenlace.
Los efectos del tratamiento se presentaron como diferencias de medias (DM) o odds ratios (OR) con intervalos de confianza (IC) del 95%. En la medida de lo posible, los datos se agruparon utilizando un modelo de efectos fijos.
Resultados principales
Esta versión actualizada comprende nueve nuevos estudios, lo que da 15 estudios incluidos y 921 participantes. La actualización también añade nuevas intervenciones, como la formación de los progenitores (dos estudios), el apoyo conductual positivo basado en la consciencia plena (o «mindfulness») (ACPBCP) (dos estudios), el entrenamiento en imitación recíproca (EIR; un estudio) y la terapia conductual dialéctica (TCD; un estudio). También añade dos nuevos estudios sobre el ACP.
La mayoría de los estudios se llevaron a cabo en la comunidad (14 estudios), y uno en un servicio forense hospitalario. Once estudios incluyeron solo a adultos. Los estudios restantes incluyeron niños (un estudio), niños y adolescentes (un estudio), adolescentes (un estudio), y adolescentes y adultos (un estudio). Un estudio incluyó a niños varones con síndrome X frágil.
Seis estudios se realizaron en Reino Unido, siete en EE. UU., uno en Canadá y uno en Alemania. Solo cinco estudios describieron las fuentes de financiación.
Cuatro estudios compararon el control de la ira basado en la terapia cognitivo‐conductual con un grupo de control en lista de espera o ningún tratamiento (n = 263); dos estudios compararon el ACP con el tratamiento habitual (n = 308); dos estudios compararon la formación de los cuidadores en consciencia plena y ACP con ACP solo (n = 128); dos estudios incluyeron formación de los progenitores en enfoques conductuales comparada con un control en lista de espera o con tratamiento habitual (n = 99); un estudio de consciencia plena la comparó con un control en lista de espera (n = 34); un estudio de terapia dialectal conductual adaptada la comparó con un control en lista de espera (n = 21); un estudio de EIR lo comparó con un control activo (n = 20) y un estudio de relajación modificada la comparó con un grupo de control activo (n = 12).
Hubo evidencia de certeza moderada de que el control de la ira podría mejorar la gravedad de la conducta agresiva después del tratamiento (DM ‐3,50; IC del 95%: ‐6,21 a ‐0,79; p = 0,01; un estudio; 158 participantes); evidencia de certeza muy baja de que podría mejorar la capacidad autoinformada de control de la ira (DM ‐8,38; IC del 95%: ‐14,05 a ‐2,71; p = 0,004; I2 = 2%; tres estudios, 212 participantes), el funcionamiento adaptativo (DM ‐21,73; IC del 95%: ‐36,44 a ‐7,02; p = 0,004; un estudio, 28 participantes) y los síntomas psiquiátricos (DM ‐0,48; IC del 95%: ‐0,79 a ‐0,17; p = 0,002; un estudio, 28 participantes) después del tratamiento; y evidencia de certeza muy baja de que no mejora la calidad de vida después del tratamiento (DM ‐5,60; IC del 95%: ‐18,11 a 6,91; p = 0,38; un estudio, 129 participantes) ni reduce el uso de servicios y los costes a los 10 meses (DM 102,99 libras esterlinas; IC del 95%: ‐117,16 a 323,14; p = 0,36; un estudio, 133 participantes).
Hubo evidencia de certeza moderada de que el ACP podría reducir la conducta agresiva después del tratamiento (DM ‐7,78; IC del 95%: ‐15,23 a ‐0,32; p = 0,04; I2 = 0%; dos estudios, 275 participantes) y evidencia de certeza baja de que probablemente no reduzca la conducta agresiva a los 12 meses (DM ‐5,20; IC del 95%: ‐13,27 a 2,87; p = 0,21; un estudio, 225 participantes). Hubo evidencia de certeza baja de que el ACP no mejora el estado mental después del tratamiento (OR 1,44; IC del 95%: 0,83 a 2,49; p = 1,21; un estudio, 214 participantes) y evidencia de certeza muy baja de que podría no reducir la utilización de los servicios a los 12 meses (DM ‐448,00 libras esterlinas; IC del 95%: ‐1660,83 a 764,83; p = 0,47; un estudio, 225 participantes).
Hubo evidencia de certeza muy baja de que el ACP podría reducir los incidentes de agresión física (DM ‐2,80; IC del 95%: ‐4,37 a ‐1,23; p < 0,001; un estudio, 34 participantes) y evidencia de certeza baja de que el ACPBCP podría no reducir los incidentes de agresión tras el tratamiento (DM ‐10,27; IC del 95%: ‐14,86 a ‐5,67; p < 0,001; I2 = 87%; dos estudios, 128 participantes).
Los motivos para disminuir la certeza de la evidencia fueron el riesgo de sesgo (en concreto, el sesgo de selección y de realización); la imprecisión (resultados de un único estudio, a menudo pequeño, IC amplios e IC que cruzan la línea de ningún efecto); y la inconsistencia (heterogeneidad estadística).
Conclusiones de los autores
Hay evidencia de certeza moderada de que los enfoques cognitivo‐conductuales como el control de la ira y el ACP podrían reducir la conducta agresiva dirigida hacia terceros a corto plazo, pero hay menos certeza acerca de la evidencia a medio y largo plazo, en particular en relación con otros desenlaces como la calidad de vida. Existe evidencia que indica que la combinación de más de una intervención podría tener beneficios acumulativos.
La mayoría de los estudios fueron pequeños y se necesitan ensayos controlados aleatorizados más amplios y sólidos, en concreto para las intervenciones en las que la certeza de la evidencia es muy baja. Se necesitan más ensayos que se centren en niños y en si las intervenciones psicológicas conllevan reducciones en el uso de medicamentos psicotrópicos.
PICO
Resumen en términos sencillos
Terapias conductuales y cognitivo‐conductuales para el tratamiento de la conducta agresiva en personas con discapacidad intelectual
Mensaje clave
‐ Las técnicas como el control de la ira (basado en la terapia cognitivo‐conductual) y el apoyo conductual positivo (ACP; basado en la terapia conductual) podrían reducir la conducta agresiva hacia terceros en personas con discapacidad intelectual (dificultades de aprendizaje).
‐ Las pruebas para otros desenlaces y terapias son menos seguras debido a que solo hay unos pocos estudios pequeños.
‐ Se necesitan más pruebas sobre qué terapias y técnicas son útiles para reducir la agresividad y mejorar otros desenlaces, como la calidad de vida.
¿En qué consiste la conducta agresiva hacia terceros y por qué es importante?
La conducta agresiva hacia terceros puede incluir agresiones físicas hacia otras personas, como golpes, patadas o lanzamiento de objetos, y daños a la propiedad. Este comportamiento puede ser una forma de comunicar que las necesidades de la persona no están cubiertas. Puede acarrear consecuencias negativas, como la exclusión de los servicios de día, el fracaso de la residencia con apoyo e ingresos inadecuados en hospitales psiquiátricos. El uso de antipsicóticos en el tratamiento de conductas desafiantes tiene una función limitada. Sin embargo, esto solo se aplica en los casos en que el riesgo es muy grave y la medicación antipsicótica solo debe ofrecerse en combinación con intervenciones psicológicas o de otro tipo. Las terapias que podrían ser útiles incluyen técnicas cognitivo‐conductuales, como el control de la ira, que ayuda a la persona a establecer vínculos entre sus pensamientos, sentimientos y conducta; y técnicas conductuales, como el apoyo conductual positivo, que pretende reducir la frecuencia del comportamiento cambiando los desencadenantes y la respuesta al comportamiento. Proporcionar intervenciones eficaces en la comunidad reduce el trauma impuesto a la persona, además de mejorar enormemente su calidad de vida. No hay evidencia sólida sobre qué técnicas son más útiles para reducir la conducta agresiva.
¿Qué se quería averiguar?
Esta es una actualización de una revisión Cochrane. El objetivo fue averiguar si técnicas como las terapias conductuales y cognitivo‐conductuales son útiles para reducir la conducta agresiva en niños y adultos con discapacidad intelectual.
¿Qué se hizo?
Para identificar los estudios pertinentes, se realizaron búsquedas en bases de datos médicas y registros de ensayos clínicos hasta marzo de 2022, así como en las listas de referencias de los artículos identificados. Dos autores de la revisión examinaron de forma independiente los títulos de los artículos para determinar su elegibilidad, extrajeron los datos y evaluaron las limitaciones de los estudios. Cuando fue necesario, se estableció contacto con los autores de los ensayos para solicitar información adicional.
¿Qué se encontró?
Se incluyeron 15 estudios que utilizaron varias terapias diferentes con un total de 921 participantes en la revisión; nueve estudios son nuevos en esta actualización. La mayoría de los estudios se realizaron en entornos comunitarios y uno en un hospital de alta seguridad. Trece estudios eran pequeños e incluían entre 12 y 63 participantes y había dos estudios grandes de 179 y 245 participantes.
Las pruebas encontradas indican lo siguiente:
‐ en comparación con una lista de espera o ningún tratamiento, es probable que el control de la ira basado en la terapia cognitivo‐conductual reduzca la conducta agresiva y posiblemente podría mejorar la capacidad para controlar la ira, la capacidad para realizar actividades cotidianas y los síntomas psiquiátricos. Sin embargo, de momento no hay pruebas suficientes sobre si mejora la calidad de vida o reduce el uso de servicios y los costes.
‐ En comparación con el tratamiento habitual, es probable que el apoyo conductual positivo reduzca la conducta agresiva, pero faltan pruebas de que mejore los síntomas de salud mental o reduzca el uso de los servicios.
‐ En comparación con una lista de espera, la consciencia plena («mindfulness») podría reducir los incidentes de agresión.
‐ En comparación con el apoyo conductual positivo solo, el apoyo conductual positivo basado en la consciencia plena podría reducir los incidentes de agresión.
¿Cuáles son las limitaciones de la evidencia?
Existe una certeza moderada de los efectos del control de la ira y del apoyo conductual positivo en la reducción de la conducta agresiva, pero menos certeza de los demás desenlaces. La mayoría de los estudios fueron muy pequeños, con solo dos estudios grandes, y aún no hay pruebas suficientes para afirmar con seguridad qué técnicas son mejores para reducir la conducta agresiva.
¿Cuál es el grado de actualización de esta evidencia?
Esta revisión está actualizada hasta marzo de 2022.
Authors' conclusions
Summary of findings
| Anger management compared to wait‐list control for outwardly aggressive behaviour in people with intellectual disability | ||||||
| Patient or population: outwardly aggressive behaviour in people with intellectual disability | ||||||
| Outcomes | Anticipated absolute effects* (95% CI) | Relative effect | № of participants | Certainty of the evidence | Comments | |
|---|---|---|---|---|---|---|
| Risk with wait‐list control | Risk with anger management | |||||
| Aggressive behaviour: severity of incidents: Aberrant Behaviour Checklist (ABC – irritability subscale – key worker report – post‐treatment (Analysis 1.2.1) | The mean aggressive behaviour: severity of incidents: Aberrant Behaviour Checklist (ABC – irritability subscale – key worker report – post‐treatment (Analysis 1.2.1) was 11.0 | MD 3.5 lower | ‐ | 158 | ⊕⊕⊕⊝ | — |
| Ability to control anger: Provocation Inventory (PI) – self‐report – post‐treatment (Analysis 1.17.1) | The mean ability to control anger: Provocation Inventory (PI) – self‐report – post‐treatment (Analysis 1.17.1) was 54.85 | MD 8.38 lower | ‐ | 212 | ⊕⊝⊝⊝ | — |
| Adaptive functioning: Adaptive Behaviour Scale – Revised, Part II (ABS‐II) (Analysis 1.26) | The mean adaptive functioning: Adaptive Behaviour Scale – Revised, Part II (ABS‐II) (Analysis 1.26) was 74.9 | MD 21.73 lower | ‐ | 28 | ⊕⊝⊝⊝ | — |
| Mental state: psychiatric symptoms measured using the Brief Symptom Inventory (BSI) (Analysis 1.27) | The mean mental state: psychiatric symptoms measured using the Brief Symptom Inventory (BSI) (Analysis 1.27) was 1.22 | MD 0.48 lower | ‐ | 28 | ⊕⊝⊝⊝ | — |
| Quality of life: Comprehensive Quality of Life Scale: Intellectual Disability (COMQoL‐ID) – post‐treatment (Analysis 1.32.1) | The mean quality of life: Comprehensive Quality of Life Scale: Intellectual Disability (COMQoL‐ID) – post‐treatment (Analysis 1.32.1) was 99.9 | MD 5.6 lower | ‐ | 129 | ⊕⊝⊝⊝ | — |
| Costs of service utilisation: Client Service Receipt Inventory (CSRI): cost per person per week of health and social care resource (in British pounds (GBP)) – follow‐up: 10 months (Analysis 1.33) | The mean costs of service utilisation: Client Service Receipt Inventory (CSRI): cost per person per week of health and social care resource (in British pounds (GBP)) – follow‐up: 10 months (Analysis 1.33) was GBP 867.09 | MD 102.99 higher | ‐ | 133 | ⊕⊝⊝⊝ | — |
| *The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). | ||||||
| GRADE Working Group grades of evidence | ||||||
| See interactive version of this table: https://gdt.gradepro.org/presentations/#/isof/isof_question_revman_web_430801803348621284. | ||||||
| a Downgraded one level due to imprecision as the results were based on one study. | ||||||
| Positive behavioural support compared to treatment as usual for people with outwardly aggressive behaviour and intellectual disability | ||||||
| Patient or population: people with outwardly aggressive behaviour and intellectual disability | ||||||
| Outcomes | Anticipated absolute effects* (95% CI) | Relative effect | № of participants | Certainty of the evidence | Comments | |
|---|---|---|---|---|---|---|
| Risk with treatment as usual | Risk with positive behavioural support | |||||
| Aggressive behaviour: Aberrant Behaviour Checklist (ABC) Total – at 6 months (Analysis 2.1.1) | The mean aggressive behaviour: Aberrant Behaviour Checklist (ABC) Total – at 6 months (Analysis 2.1.1) was 56.53 | MD 7.78 lower | ‐ | 275 | ⊕⊕⊕⊝ | — |
| Ability to control anger ‐ not measured | ‐ | ‐ | ‐ | ‐ | ‐ | — |
| Adaptive functioning ‐ not measured | ‐ | ‐ | ‐ | ‐ | ‐ | — |
| Mental state: Common Mental Disorder (CMD) – at 6 months (Analysis 2.2.1) (CMD) | 368 per 1000 | 456 per 1000 | OR 1.44 | 215 | ⊕⊕⊝⊝ | — |
| Quality of life ‐ not measured | ‐ | ‐ | ‐ | ‐ | ‐ | — |
| Costs of service utilisation: mean health and social care costs per participant – at 12 months (Analysis 2.7.2) (Cost of service utilisation) | The mean costs of service utilisation: mean health and social care costs per participant – at 12 months (Analysis 2.7.2) was GBP 4051 | MD 448 lower | ‐ | 225 | ⊕⊝⊝⊝ | — |
| *The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). | ||||||
| GRADE Working Group grades of evidence | ||||||
| See interactive version of this table: https://gdt.gradepro.org/presentations/#/isof/isof_question_revman_web_430802235954903680. | ||||||
| a Downgraded one level due to imprecision as confidence intervals were wide. | ||||||
| Meditation based on mindfulness compared to treatment as usual for people with outwardly aggressive behaviour and intellectual disability | ||||||
| Patient or population: people with outwardly aggressive behaviour and intellectual disability | ||||||
| Outcomes | Anticipated absolute effects* (95% CI) | Relative effect | № of participants | Certainty of the evidence | Comments | |
|---|---|---|---|---|---|---|
| Risk with treatment as usual | Risk with meditation based on mindfulness | |||||
| Aggressive behaviour: frequency of incidents: number of incidents of physical aggression per week during treatment (12 weeks) – post‐treatment (Analysis 3.1) | The mean aggressive behaviour: frequency of incidents: number of incidents of physical aggression per week during treatment (12 weeks) – post‐treatment (Analysis 3.1) was 5.80 | MD 2.8 lower | ‐ | 34 | ⊕⊝⊝⊝ | — |
| Ability to control anger ‐ not measured | ‐ | ‐ | ‐ | ‐ | ‐ | No studies were available |
| Adaptive functioning ‐ not measured | ‐ | ‐ | ‐ | ‐ | ‐ | No studies were available |
| Mental state ‐ not measured | ‐ | ‐ | ‐ | ‐ | ‐ | No studies were available |
| Quality of life ‐ not measured | ‐ | ‐ | ‐ | ‐ | ‐ | No studies were available |
| Cost of service utilisation ‐ not measured | ‐ | ‐ | ‐ | ‐ | ‐ | No studies were available |
| *The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). | ||||||
| GRADE Working Group grades of evidence | ||||||
| See interactive version of this table: https://gdt.gradepro.org/presentations/#/isof/isof_question_revman_web_430895877541611244. | ||||||
| a The evidence was downgraded three levels due to risk of bias from study limitations (unclear risk of selection bias and high risk of detection bias due to lack of blinding of study assessors) and imprecision (results from one study with small sample size and wide confidence intervals). | ||||||
| Mindfulness based on positive behaviour support compared to positive behaviour support in outwardly aggressive behaviour in people with intellectual disability | ||||||
| Patient or population: outwardly aggressive behaviour in people with intellectual disability | ||||||
| Outcomes | Anticipated absolute effects* (95% CI) | Relative effect | № of participants | Certainty of the evidence | Comments | |
|---|---|---|---|---|---|---|
| Risk with positive behaviour support | Risk with mindfulness based on positive behaviour support | |||||
| Aggressive behaviour: aggressive events – carer reports – post‐training (Analysis 4.1) | The mean aggressive behaviour: aggressive events – carer reports – post‐training (Analysis 4.1) was 15.78 | MD 10.27 lower | ‐ | 128 | ⊕⊕⊝⊝ | — |
| Ability to control anger ‐ not measured | ‐ | ‐ | ‐ | ‐ | ‐ | — |
| Adaptive functioning ‐ not measured | ‐ | ‐ | ‐ | ‐ | ‐ | — |
| Mental state ‐ not measured | ‐ | ‐ | ‐ | ‐ | ‐ | — |
| Quality of life ‐ not measured | ‐ | ‐ | ‐ | ‐ | ‐ | — |
| Cost of service utilisation ‐ not measured | ‐ | ‐ | ‐ | ‐ | ‐ | — |
| *The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). | ||||||
| GRADE Working Group grades of evidence | ||||||
| See interactive version of this table: https://gdt.gradepro.org/presentations/#/isof/isof_question_revman_web_430892291844998974. | ||||||
| a Downgraded two levels due to risk of bias arising from study limitations (lack of blinding of outcome assessors) and inconsistency arising from heterogeneity. | ||||||
Background
Description of the condition
Intellectual disabilities
Intellectual disabilities (ID), also known as 'learning disabilities' in the UK, is defined as a condition of global cognitive delay that occurs during the developmental period. It is characterised by a low cognitive ability as indicated, for example, by an intelligence quotient (IQ) below 70 on an appropriately standardised and administered test. It is also characterised by significant additional deficits in a range of areas of adaptive functioning such as education, occupation, self‐direction, personal relationships and community utilisation. The term 'disorders of intellectual development' replaces 'mental retardation' in the International Classification of Diseases, 11th Revision (ICD‐11) (WHO 2019) and 'intellectual disability' is the term used in the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM‐5) (APA 2013). We use the term 'intellectual disabilities' throughout the present review.
Problem behaviours in people with intellectual disabilities
Problem (also called challenging) behaviour is a term used to describe longstanding patterns of maladaptive behaviour in people with IDs. While not currently a diagnostic entity, it is nonetheless considered a potential comorbidity in this population in the DSM‐5 (APA 2013). In ICD‐11 there is a new diagnostic domain, disruptive behaviour and dissocial disorders, which is now considered to occur across the lifespan. There is acceptance that a social perspective on challenging behaviour is very important because it is the complex interaction between the individual and the setting in which they live that partly determines whether a behaviour is considered unacceptable. It will, for example, have to be interpreted as frightening or distressing by another person (RCPsych 2007). Emerson 1995 emphasised this 'two‐way' interaction between the individual and their environment in his definition of challenging behaviour: (quote) "It refers to culturally abnormal behaviour of such intensity, frequency or duration that the physical safety of the person or others is likely to be placed in serious jeopardy, or behaviour which is likely to seriously limit the use of, or result in the person being denied access to ordinary community facilities".
In clinical practice in the UK, the consensus Diagnostic Criteria for Behavioural and Mental Disorders in Learning Disabilities (DC‐LD) have adopted the following general diagnostic criteria for problem behaviours in people with IDs (RCPsych 2001):
-
A. The problem behaviour is of significant frequency, severity or chronicity as to require clinical assessments and specialist interventions or support;
-
B. The problem behaviour must not be a direct consequence of other psychiatric disorders (e.g. pervasive developmental disorders, non‐affective psychotic disorders, depressive episode, generalised anxiety disorders, personality disorders, drugs or physical disorders);
-
C. One of the following must be present:
-
the problem behaviour results in significant negative impact on the person's quality of life or that of others; or
-
the problem behaviour presents significant risks to the health or safety of the person or others.
-
-
D. The problem behaviour is persistent and pervasive.
Throughout this systematic review, we use the term "aggressive challenging behaviour," which we conceptualise as "emotional or impulsive aggression" which refers to "aggression that occurs with only a small amount of forethought or intent and that is determined primarily by impulsive emotions" (Jhangiani 2014).
Aggressive challenging behaviour includes:
-
physically or verbally aggressive behaviour;
-
destruction of environment;
-
sexually inappropriate behaviour and
-
offending type behaviour (arson, stealing, other crime) (RCPsych 2001).
Several types of aggressive behaviour may be displayed by the same individual (Borthwick‐Duffy 1994). Severity correlates with level of IQ, gender, institutional setting, age and other disabilities.
One US survey of 91,164 people with IDs identified aggressive challenging behaviours (physical aggression and property destruction) in 9.2% of all those who presented with any type of problem behaviours (14%) (Borthwick‐Duffy 1994). One study of 3165 adults with IDs receiving rehabilitation services in Canada found a 12‐month prevalence of aggressive behaviour of 51.8%; property damage in 24%; verbal aggression of 37.6%; physical aggression of 24.4% and sexual aggression of 9.8% (Crocker 2006). However, only 4.9% of individuals exhibited aggressive behaviour that led to another person being injured. One community‐based epidemiological study found that the prevalence of aggression was 9.8% and the two‐year incidence was 1.8% (Cooper 2009). This was replicated in a more recent total population survey in England where there was recorded aggressive challenging behaviour in 8.3% of participants (Bowring 2017). The prevalence of aggression in individuals with profound ID or multiple disabilities has been reported to be as high as 45% (Poppes 2010).
McClintock 2003 identified several predictors for aggressive challenging behaviour in people with IDs in one meta‐analytic study, such as male gender, deficit in expressive communication and a diagnosis of autism. There is indication from a few studies that there may be a gender association with aggressive challenging behaviour in that females are more likely to display verbal aggression whilst males are more likely to display any type of aggressive challenging behaviour (NICE 2015).
The chronic remitting‐relapsing course of aggressive challenging behaviour and its contribution to social exclusion, application of restrictive practices, and poor quality of life and reduced life expectancy are strong reminders for the need for immediate and accessible specialised support and for the development of effective interventions for its management (Lundgvist 2013).
Scope of this review
This review is concerned with behavioural and cognitive‐behavioural interventions for outwardly directed aggressive behaviour, which refers to aggressive and destructive behaviour directed towards others and property, including verbal and physical aggression. The review does not include studies of self‐injurious behaviour, examined in another Cochrane Review (Rana 2013), or other types of behaviour such as stereotypical behaviour or sexually inappropriate behaviour.
There are two main approaches to treating and managing aggressive challenging behaviour which however have little specificity to the actual presentation: cognitive behavioural therapy (CBT)‐based anger management and positive behaviour support (PBS). However, there several other psychological interventions that have been adapted and tested, albeit in small uncontrolled trials, such as dialectical behaviour therapy (DBT) and mindfulness but none of these interventions are fully rolled out in clinical settings beyond PBS which is considered first‐line treatment. However, as aggressive challenging behaviour is a complex presentation with many underlying reasons including genetic, psychological and social, there is no single intervention that can be appropriate or suitable for all. Therefore, a range of psychosocial interventions are required to address this condition. Often, psychosocial interventions may be delivered alongside pharmacological treatments (e.g. antipsychotics), and multimodal approaches may be more effective given the frequent presence of comorbid disorders such as attention deficit hyperactivity disorder and autism. For more information about the overall management of challenging behaviour including aggression, see Ali 2014 and Hassiotis 2022.
Description of the intervention
Management of aggressive behaviour
Established patterns of aggressive behaviour usually require intensive and individualised management. Generally, the acute response to severe aggression that may pose risk to others requires de‐escalation and possibly the removal of the individual from the vicinity by implementing techniques, which may include appropriate forms of restraint, time‐out or psychotropic medication. Psychosocial techniques can then be implemented to promote more appropriate behavioural responses and to prevent further episodes.
A range of psychosocial interventions, which contain behavioural and cognitive‐behavioural techniques, are available to manage aggressive behaviour in people with IDs. While we make a distinction between behavioural and cognitive‐behavioural interventions, it may be a false dichotomy to think of them as unconnected methodologies. Most current interventions integrate elements stemming from both cognitivism and behaviourism to elicit change, but may differ in levels of emphasis of the two within what is often a complex intervention (Skivington 2021).
Behavioural approaches aim to change maladaptive behaviours by reinforcing desirable behaviours and reducing or eliminating unwanted behaviours. Such approaches may include behavioural analysis, which involves carrying out a functional assessment to determine the function or purpose of the behaviour (Iwata 1994). This may include the desire for tangible goods, the need to avoid or escape a situation, the need for social attention or it may be self‐stimulatory, in the case of repetitive behaviours seen in the context of autism. A clear description of the behaviour, factors that predict the behaviour, and factors or consequences that maintain it are identified. A hypothesis about the function of the behaviour is formulated, followed by direct observations that provide evidence supporting the hypothesis and implementation of strategies aimed at modifying the antecedents or consequences of the behaviour. Behavioural approaches that aim to understand the function of problem behaviours include functional communication training (FCT) and non‐contingent reinforcement.
Some behavioural interventions are based on a framework of tiered interventions addressing aggressive challenging behaviour of increasing severity and frequency (Tincani 2020). PBS is one framework used frequently in the UK (Allen 2005). It aims to encourage a clear understanding of the individual's behaviour within their social context. Gore 2022 have developed an updated model and conceptualisation of PBS since their earlier paper (Gore 2013). They propose 12 core components of PBS covering 'rights and values' (three components), 'theory and evidence base' (three components) and 'process and strategy' (four components). The theoretical framework underpinning behaviours that challenge is the biopsychosocial model, which suggests that a range of factors, such as communication difficulties, mental and physical health problems, sensory needs, trauma, and social and environmental factors, may contribute to challenging behaviours. As well as influencing change at the individual level, the model advocates for changes at a broader, systems or population level (whole systems approach) that attempts to reduce exposure to adversity and to increase the quality of support provided for people with ID, thus preventing behaviours that challenge. Therefore, approaches can be directed at the 'primary level' where the focus is on ensuring that individuals have a good quality of life; at the secondary level the emphasis is on early identification and support for those at risk of challenging behaviours, and the tertiary level, where highly specialised support is provided for those at greatest risk of challenging behaviours. Gore 2022 emphasise the importance of working in partnership with the person with ID and their carers, multidisciplinary working, the need for high‐quality care and support environments (e.g. 'capable environments' McGill 2020) and conducting a bespoke functional assessment.
Reciprocal imitation training (RIT) is a naturalistic developmental behavioural intervention that was developed for children with severe IDs, autism or both (Ingersoll 2009) where there may be an underlying deficit of imitation skills, which are crucial in the development of social skills and communication. It involves two individuals alternating between imitating and being imitated by the other person.
CBT, a widely used approach for a range of psychological problems, aims to allow the individual to identify how their thoughts, feelings and physiological responses are linked, to allow them to address present maladaptive cognitions (Howton 1989). The therapist and individual develop a formulation of the causes behind the behaviours and factors that maintain it, and work together to identify new strategies to reduce distress and cope with difficult situations. The therapist helps the patient learn to recognise errors in thinking style and explore potential alternative explanations for their beliefs. The therapist makes necessary technical modifications based on the level of ID and the individual needs of their client, for example, visual aids, repetition or proceeding at a slower pace. Such adapted CBT has been used mainly for the management of aggression and anger in people with IDs. DBT uses both CBT and behavioural approaches and was first developed by Marsha Linehan for the treatment of individuals with borderline personality disorder who frequently experience emotional dysregulation (Linehan 1987; Linehan 1993). It has now been used in the treatment of a range of mental disorders, as well as the treatment of anger and aggression (Ciesinski 2022). The core components of DBT are weekly individual sessions, a weekly group skills training session, a therapist consultation team meeting and telephone crisis coaching.
Other interventions such as mindfulness, which may incorporate elements from other therapies such as PBS, have also been tested to support people with IDs who display aggressive challenging behaviour or outwardly directed aggression. Mindfulness has its roots in key principles found in Buddhist teaching (Bishop 2004), and utilises meditation to identify patterns in thinking. Patients learn to manage their emotional reactions by observing anger‐related sensations in the body without attempting to avoid or act on them.
In practice, there is often a multicomponent approach to the management of aggressive behaviour, as it is important to bring about significant changes rapidly and to ensure the safety of the person and others, including carers. However, the objective is to promote long‐term emotional well‐being and reduce challenging behaviours.
How the intervention might work
FCT involves identifying the function or reinforcer of the problem behaviour and then teaching a functionally equivalent communicative response to replace the problem behaviour (Fisher 1993). Non‐contingent reinforcement involves the response independent delivery of reinforcers. The association between the behaviour and the reinforcer is reduced or eliminated by providing the reinforcer at fixed or variable times, thus attenuating the motivation for the problem behaviour (Carr 2009), and may be accompanied by reinforcement of more appropriate behaviour.
At the individual (tertiary) level, PBS involves using functional analysis to identify the purpose of a particular behaviour, and then teaching the person the necessary social, behavioural and communication skills to replace the original behaviour (Allen 2005). The approach demands a multicomponent intervention as it acknowledges that challenging behaviours may be provoked by multiple variables. Therefore, it involves the development of a new support plan and focuses on achieving physical, environmental and social well‐being for the individual. It relies on the use of positive strategies and reinforcement to maintain positive behaviour, avoiding any punitive measures.
CBT‐based interventions for anger management can be delivered either individually or within a group setting. In order to successfully implement the intervention, the therapist and individual work collaboratively to identify and develop awareness of situations that evoke anger in the individual and when they are becoming angry. It teaches new skills to manage and control anger when such situations arise. Anger management strategies involve a range of cognitive, behavioural and physiological skills such as problem‐solving, removing oneself from the situation, being assertive, seeking help or employing relaxation techniques.
Mindfulness teaches the individual to focus and refocus their attention on the here and now during meditation and to observe their thoughts and feelings without judgement (Bishop 2004). In practice this means realising when they have feelings of anger or upset without trying to censor or analyse these feelings or thoughts. Individuals learn to develop the ability to let such feelings arise and subside without having to react to them, thus changing patterns of behaviour. An intervention based on mindfulness called 'Meditation on the Soles of the Feet' teaches individuals with aggressive behaviour to identify behaviours or emotions that give rise to anger or aggression and then to alter the focus of their attention from these precursors to a neutral part of their body, in this case, the soles of the feet (Singh 2003).
DBT teaches individuals four core skills that include mindfulness, interpersonal effectiveness (setting clear goals, maintaining self‐respect and reducing conflicts in relationships), distress tolerance and managing emotional dysregulation (Linehan 1987).
RIT teaches the individual to learn imitation skills that enable the development of more appropriate behaviours, thereby potentially reducing challenging behaviours (Ingersoll 2009).
Why it is important to do this review
Policy and clinical context
There has been growing concern in the UK over the negative consequences of problem behaviours displayed by individuals with IDs, which include breakdown in accommodation and placements (Phillips 2010), admission into specialist inpatient units (Oxley 2013) and exposure to restrictive practices such as restraint and seclusion (McGill 2009). Since the Winterbourne View scandal that unveiled widespread emotional and physical abuse of inpatients at Winterbourne View Hospital (Department of Health 2012), government policies such as Transforming Care (NHS England 2015a; NHS England 2015b) aimed to improve the care and treatment of individuals displaying problem behaviours, including outwardly directed aggressive behaviour, with a view to enhancing the provision of specialist community services and reducing inpatient admissions. In addition, there is the STOMP (Stopping Over‐Medication of People with learning disability, autism or both) UK national project that aims to reduce the prescribing of unnecessary psychotropic medication in people with ID and challenging behaviours who do not have a mental health diagnosis (NHS England 2019). In order to achieve this objective, there is a need for alternative non‐pharmacological interventions. Therefore, this updated review is important given this wider context. Reviewing the evidence base of the types of interventions that might be effective in managing individuals with outwardly aggressive behaviour may help to improve access and commissioning of effective interventions in community intellectual services, thus preventing unnecessary hospital admissions.
Previous research
The importance of effectively managing problem behaviours, particularly aggression, has led to considerable research into different types of interventions. These are mostly single‐case reports, which lack data on long‐term evaluation of clinical and cost‐effectiveness.
Scotti 1991 suggested that formal functional analysis of behaviour prior to treatment may improve clinical outcomes.
Didden 1997 included four groups of interventions: pharmacological management, antecedent control procedures, response contingent procedures and non‐contingent procedures. The participants had a diagnosis of severe ID or developmental disorders such as autism. The authors used 'elimination of problem behaviour' as their main outcome. They compared pre‐ and post‐treatment effects and concluded that most problem behaviours could be treated with variable degrees of success. Aggressive behaviour (aggression, destruction, public disrobing) was less amenable to treatment when compared to self‐injury or socially disruptive behaviours such as tongue protrusion, non‐compliance and hyperactivity. The authors concluded that externally destructive behaviours "are associated with the lowest levels of treatment efficacy".
Carr 1999 conducted a review of studies using PBS and identified 107 papers (single‐case studies). PBS was widely applicable to individuals with developmental disabilities with serious problem behaviour and PBS was effective in reducing problem behaviours in 50% to 66% of cases. The success rates were higher when a prior functional assessment was completed.
Didden 2006 conducted a meta‐analysis of single‐case studies of individuals with mild IDs to borderline intelligence. The authors concluded that behavioural interventions based on functional behavioural analysis pretreatment could improve several types of problem behaviours, including destructive behaviour, physical and verbal aggression, self‐injury and stereotypies.
Heyvaert 2012 conducted a multilevel meta‐analysis examining 285 single‐case and small studies for the effectiveness of contextual, behavioural and pharmacological interventions for problem behaviours in people with IDs. Contrary to previous meta‐analyses, the authors did not find pretreatment functional assessment to be a significant moderator. Only aggression and the manipulation of antecedent factors had significant moderating effects. Outwardly directed aggression appeared to be associated with significantly lower effectiveness of interventions for problem behaviours, while the presence of a component manipulating antecedent factor, such as environmental or social factors, was associated with a better outcome than when no such component was present.
Another systematic review and meta‐analysis of the effectiveness of CBT in the management of anger included 12 studies (Nicoll 2013). Nine studies were suitable for meta‐analysis. The authors found that CBT treatment had large effect sizes, but suggested that this should be regarded with caution due to the inclusion of studies with small sample sizes. They included non‐randomised and uncontrolled studies, which are likely to introduce bias.
One meta‐analysis using single‐case design studies of non‐contingent reinforcers for problem behaviours (including aggression) in people with developmental disorders (over 50% had ID) was carried out by Richman 2015. They included 55 studies with 91 participants and found that the intervention had a large effect size for reducing problem behaviours.
Gerow 2018 identified 215 single‐case design studies that evaluated FCT in individuals with a range of disabilities including ID and autism. Individuals had challenging behaviours including disruptive and aggressive behaviour. They found 135 studies demonstrating moderate‐to‐strong evidence for the use of FCT in reducing challenging behaviours.
All of these reviews suggest that CBT and behavioural approaches may be at least partially effective in the management of aggressive challenging behaviour. However, these reviews had methodological shortcomings such as the inclusion of uncontrolled studies. We consider that there is a need for the present systematic review as more trials have been conducted in the field of intellectual and developmental disorders since the last review and, therefore, an update was necessary to reflect the current state of science and evidence in the field.
Objectives
To evaluate the efficacy of behavioural and cognitive‐behavioural interventions on outwardly directed aggressive behaviour compared to usual care, wait‐list controls or no treatment in people with intellectual disability. We also evaluated enhanced interventions compared to non‐enhanced interventions.
Methods
Criteria for considering studies for this review
Types of studies
We included randomised controlled trials (RCTs), such as individual and cluster‐RCTs, and also allowed for other variants (e.g. stepped wedge designs). We included quasi‐RCTs where participants were allocated to different arms of the study but the allocation was not truly random (e.g. every other participant was allocated to the intervention arm or allocation was based on date of birth).
Types of participants
Children and adults with IDs (mild‐to‐severe/profound) who exhibited aggressive behaviour were the focus of this review. We considered studies of participants with pervasive developmental disorders, such as autism, if they stated that the participants also met criteria for IDs by some standardised measure or were recorded as having been previously assessed (APA 2000; WHO 2019). We included studies where participants had other comorbid conditions in addition to IDs and aggressive behaviour if it was possible to extract data on aggressive behaviour distinct from other symptoms. We specified that at least 70% of the sample should comprise participants with IDs if other populations were included in the study. We excluded studies where the participants had an adult‐onset organic brain disorder such as dementia.
As in the previous update, we include syndromes associated with an aggressive behavioural phenotype such as Prader‐Willi syndrome, Williams syndrome, Fragile X syndrome and tuberous sclerosis, as studies of interventions targeting these behaviours may have been completed and published in the interim. Our rationale for including these syndromes was that people with aggressive behavioural phenotypes are often the most difficult to treat, and the evidence base for the management of aggressive behaviour in these groups is limited. Including these participants in the review increases the utility of the results, and should help clinicians to make better decisions about the treatment of people with aggressive behaviour and IDs, including those with behavioural phenotypes.
Types of interventions
The interventions examined were as follows.
-
Behavioural interventions. This was applied broadly to include interventions aimed at individuals to reduce the problem behaviour. Interventions that were considered appropriate for inclusion in the review were those based on the principles of applied behavioural analysis such as non‐contingent reinforcement, FCT, PBS and parenting interventions using a behavioural approach. We also included RIT, which is not based on traditional principles of applied behavioural analysis but is described as a 'naturalistic behavioural intervention' (Ingersoll 2009).
-
CBT‐based interventions. Interventions considered in this category were anger management, problem‐solving skills training, relaxation training, meditation, mindfulness and DBT.
We included interventions that used both approaches (e.g. mindfulness‐based positive behaviour support (MBPBS)). We did not restrict interventions to specific settings but covered hospitals, community day centres and individuals' own homes. We included individual and group treatments. We included interventions delivered to parents and carers if they were involved in the delivery of the intervention to individuals with IDs.
For the purpose of this review, we excluded sensory‐based interventions and sensory integration therapy (SIT). SIT uses play‐based sensory motor activities to help the individual process sensory information and focuses on the therapist–child/patient relationship (Schaaf 2015). It is usually delivered by occupational therapists and is usually aimed at individuals with autism, who are more likely to have difficulties with sensory processing. Although it is thought to improve behaviours that challenge, it is not described as a 'behavioural intervention' and, therefore, was excluded.
The control groups:
-
received standard care or treatment as usual (TAU), that is, general physical and psychological care but no specific intervention programme for the expressed aim of treating the behaviour;
-
compared a standard method of intervention against a more specialised or enhanced version;
-
were active control group (participants were offered another activity);
-
were on a waiting list or
-
received no treatment.
Types of outcome measures
Studies had to include at least one of the following primary outcomes, measured using a standardised instrument (Aman 1991), or based on direct observations or recordings provided that there was adequate reporting of the reliability of data. Changes could be reported by individuals, family or paid carers.
Primary outcomes
-
Change in aggressive behaviour (frequency/severity of incidents), for example measured using the Aberrant Behavior Checklist (Aman 1985), or number of incidents (Singh 2013; Singh 2016; Singh 2020a)
-
Change in ability to control anger, for example measured using the Provocation Inventory (PI) (Novaco 2003)
-
Change in adaptive functioning, for example using the Matson Evaluation of Social Skills for Individuals with Severe Mental Retardation (MESSIER; Matson 1998)
-
Adverse effects, such as death, or adverse effects from treatment
Secondary outcomes
-
Change in mental state, for example measured using the Mini Psychiatric Assessment Schedules for Adults with Developmental Disabilities (Mini PAS‐ADD) (Prosser 1998)
-
Change in (additional) medication
-
Change in care needs
-
Change in quality of life, for example measured using the Comprehensive Quality of Life Scale: Intellectual Disability (COMQoL‐ID) (Cummins 1997)
-
Frequency of service utilisation (and costs if available), for example measured using the Client Service Receipt Inventory (CSRI) (Beecham 1992)
-
User satisfaction data (or dissatisfaction if known)
Timing of outcome measures
We categorised the follow‐up periods as medium‐term (three to six months) and long‐term (nine months or longer) from the end of treatment.
Search methods for identification of studies
In this updated review, we revised the search terms to include PBS. We ran searches for all available years for each database, to reduce the likelihood of missing studies, and deduplicated these records with those from previous searches. We modified the search terms where necessary for each of the databases searched. We ran searches for this update in August 2020 applying no date or language limits. We conducted a top‐up search in March 2022 to identify studies published since August 2020.
Electronic searches
-
Cochrane Central Register of Controlled Studies (CENTRAL; 2022, Issue 2) in the Cochrane Library (searched 21 March 2022)
-
MEDLINE Ovid (1946 to March week 2 2022)
-
MEDLINE In‐Process and Other Non‐indexed Citations Ovid (18 March 2022)
-
MEDLINE Epub Ahead of Print Ovid (18 March 2022)
-
Embase Ovid (1974 to 2022, March 16)
-
APA PsycINFO Ovid (1806 to March week 2 2022)
-
CINAHL Plus EBSCOhost (1937 to 21 March 2022)
-
Cochrane Database of Systematic Reviews (CDSR; 2022, Issue 3) in the Cochrane Library (searched 21 March 2022)
-
ERIC EBSCOhost (1966 to 22 March 2022)
-
ProQuest Dissertations & Theses Abstracts Global (searched 22 March 2022)
-
Web of Science Core Collection Clarivate (Science Citation Index, Social Science Citation Index, Conference Proceedings Citation Index –Science, Conference Proceedings Citation Index – Social Sciences & Humanities, Emerging Sources Citation Index) (searched 22 March 2022)
-
World Health Organization International Clinical Trials Registry Platform (trialsearch.who.int/) (searched 22 March 2022).
-
ClinicalTrials.gov (www.clinicaltrials.gov/) (searched 22 March 2022).
The search strategies for this update are reported in Appendix 1. The search strategies for previous versions of this review are reported in Hassiotis 2004, Hassiotis 2008, and Ali 2015.
Searching other resources
We scrutinised the reference lists of relevant studies retrieved through the electronic searches.
Data collection and analysis
Selection of studies
Following the first search, two review authors (DP and SW) independently screened titles and abstracts of all records returned by the search. The same two review authors then retrieved the full text of papers that appeared relevant, or for which we needed more information, and assessed them for eligibility. Following the top‐up search, two review authors (DP and MM) completed the above steps. For studies where there was a difference of opinion, we reached a consensus through discussion with the other review authors (AH, AA and IH). We recorded the flow of studies through this process in a PRISMA flow diagram which describes the screening process and reasons for excluding studies (Figure 1) (Moher 2009).
Study flow diagram.
If a review author was also a contributor to a study that may have been of interest to the review, that review author was not involved in the study selection process.
Data extraction and management
We developed a data extraction form for this review adapted from the that used in the previous review. Teams of two review authors (from DP, SW and MM) independently undertook data extraction for each of the included reports. Extracted information included study location, methods, participant details, type of intervention, intensity and duration of intervention, outcomes and information to allow for risk of bias judgements. Where there were gaps in the available data, we attempted to contact the study authors for further information (e.g. where a study claimed to be randomised but gave little detail of the exact process). Where there was a difference of opinion, we reached a consensus with the other review authors (AH, AA and IH).
If a review author was also a contributor to a study that may have been of interest to the review, that review author was not involved in the data extraction process.
Assessment of risk of bias in included studies
Teams of two review authors (from DP, SW, MM) independently assessed each paper using the Cochrane RoB 1 tool for assessing risk of bias (Higgins 2011), which covers the following seven domains. Where there was a difference of opinion, we reached a consensus with the other review authors (AH, AA and IH).
-
Random sequence generation (selection bias), which we judged according to whether there was evidence of an adequate generation of a randomisation sequence.
-
Allocation concealment (selection bias), which we judged according to whether there was a detailed description of the method used to conceal the allocation sequence in order to prevent the intervention allocation from being predicted or foreseen in advance of or during enrolment into the study.
-
Blinding of participants and personnel (performance bias), which we assessed according to whether the participants or personnel involved in the study were aware of which intervention the participants received.
-
Blinding of outcome assessment (detection bias), which we assessed according to whether the outcome assessors were blinded to the allocation group.
-
Incomplete outcome data (attrition bias), which we assessed by the presence of incomplete outcome data.
-
Selective outcome reporting (reporting bias), which we assessed by examining whether the study reported all prespecified outcomes.
-
Other sources of bias, which we assessed by recording any noteworthy concerns not addressed in the above domains. For example, risk of bias due to concerns about treatment fidelity, or if the authors demonstrated strong allegiance to the intervention.
For each included study, we assigned a judgement of unclear, low or high risk of bias to each of the seven domains according to the criteria outlined in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). See Characteristics of included studies table and accompanying risk of bias tables.
If a review author was also a contributor to a study that may have been of interest to the review, that review author was not involved in the risk of bias assessment process.
Measures of treatment effect
Dichotomous data
For binary outcomes, for example, problem behaviour 'present' or 'not present', we calculated a standard estimation of the odds ratio (OR) with 95% confidence intervals (CI).
Continuous data
We calculated the mean difference (MD) if the mean and standard deviations (SD) were available or could be extrapolated from the data using test statistics (standard errors, 95% CIs, P values). Where the data were not reported in the manuscript, we contacted authors and asked them to provide this information.
See Appendix 2 on how we plan to analyse data if studies use different measures to assess the same outcome in future updates of this review.
Unit of analysis issues
The most common unit of analysis was the individual participant. The cluster‐RCTs provided individual data that was adjusted in the analysis for clustering by centre (e.g. multilevel model). See Appendix 2 for our planned approach if clustering is not taken into account in the study in future updates of this review (Higgins 2022).
Dealing with missing data
We assessed the missing data for each included study and reported the numbers of dropouts in the risk of bias tables in the Characteristics of included studies table. We attempted to contact study authors to obtain any missing data. If we were unable to impute the data (e.g. we had insufficient information regarding the numbers missing from each group), we analysed the available data only and provided an explanation in the main text the reasons why we were unable to impute the missing data. When we were unable to impute the results, we presented a narrative description of the results of these studies. See Appendix 2 for additional methods for dealing with missing data archived for future updates of this review.
Assessment of heterogeneity
We assessed clinical heterogeneity by examining differences in the characteristics of the samples, methodology, interventions and types of outcome measures used.
We assessed statistical heterogeneity studies pooled in a meta‐analysis, which was only possible for three comparisons. We inspected statistical heterogeneity by visual inspection of the forest plots to examine the extent to which CIs overlapped around the estimate for each included study on the forest plots. If CIs had generally limited overlap, this indicated the presence of statistical heterogeneity. We used the Chi² test to formally test for heterogeneity, with P less than 0.10 indicating possible heterogeneity (Deeks 2022). We used the I² statistic to determine the proportion of variation in point estimates that was due to heterogeneity rather than sampling error or chance. We considered I² values in the range of 30% to 60% to indicate moderate heterogeneity; 50% to 90% as substantial heterogeneity and 75% to 100% as considerable heterogeneity (Deeks 2022).
Assessment of reporting biases
The impact of reporting biases was reduced by our undertaking of comprehensive searches of multiple sources and identifying whether all outcomes had been reported by the studies.
Due to the small number of studies included in the review, we were unable to perform funnel plots to assess reporting bias. For information on how we plan to assess reporting biases in future updates of the review, see Appendix 2.
Data synthesis
We undertook separate comparisons of each intervention category comparing outcomes between the intervention and control groups at post‐treatment and follow‐up where available. Where studies were sufficiently similar, we pooled the data in a meta‐analysis using a fixed‐effect model, which assumes that studies are estimating the same (fixed) treatment effects. If there was variability in the population and interventions and considerable statistical heterogeneity (I2 values of 75% and above), we analysed the data using a random‐effects model. Meta‐analysis was only possible for three comparisons. The analysis was conducted using Review Manager Web (RevMan Web 2022). For the studies we were unable to pool, we presented a narrative description of the study findings.
Subgroup analysis and investigation of heterogeneity
We could not conduct the planned subgroup analysis in this review update. For information on how we plan to conduct subgroup analysis in future reviews, see Appendix 2.
Sensitivity analysis
We did not conduct any sensitivity analysis in this review. See Appendix 2.
Summary of findings and assessment of the certainty of the evidence
We presented the main findings of the review in summary of findings tables, which we created for the main interventions and comparisons using GRADEpro GDT (GRADEpro GDT). We selected the following outcomes for inclusion in the tables.
-
Change in aggressive behaviour (frequency/severity of incidents)
-
Change in ability to control anger
-
Changes in adaptive functioning
-
Change in mental state
-
Change in quality of life
-
Frequency and costs of service utilisation
Two review authors (DP and AA) assessed the certainty of evidence for each outcome using the GRADE approach, resolving any discrepancies by discussion (Schünemann 2022). We rated the certainty of evidence as high, moderate, low or very low according to the following five criteria: risk of bias in study designs (e.g. presence of limitations in design and implementation); indirectness of evidence (e.g. whether the study populations in the studies were comparable to the population of interest); heterogeneity or inconsistency; imprecision of the effect estimates and risk of publication bias. The evidence was downgraded one level (e.g. to moderate certainty) if there were serious concerns in one of these domains; evidence was downgraded two levels (e.g. low certainty) if there were serious concerns in two domains or very serious concerns in one domain; and evidence was downgraded three levels (e.g. very low certainty) if there were serious concerns in three domains or very serious concerns in two domains (or combinations of serious and very serious concerns) or extremely serious concerns in one domain. The risk of bias assessment was based on whether study limitations were likely to reduce the confidence in the effect for that particular outcome. If the risk of bias from study limitations was low (e.g. low risk of bias across all domains), the evidence was not downgraded. If there was one crucial limitation (e.g. high risk of bias) or several minor limitations (e.g. unclear risk of bias), the evidence was downgraded one level; if there were two crucial limitations (two domains with high risk of bias) or one domain with high risk of bias and several minor limitations (unclear risk of bias), the evidence was downgraded two levels. Blinding of participants was not included in the criteria due to the nature of the intervention. The certainty of the evidence was described in the summary of findings tables and the rationale for downgrading the evidence was listed in the footnotes. Due to the large number of comparisons, we selected the following five comparisons for inclusion within the summary of findings tables.
-
Anger management compared to wait‐list control
-
Positive behavioural support compared to TAU
-
Setting‐wide positive behavioural support (social care improvement plan and PBS) compared to positive behavioural support
-
Meditation based on mindfulness compared to TAU
-
Mindfulness based on positive behavioural support compared to positive behavioural support
If a review author was also a contributor to a study that may have been of interest to the review, that review author was not involved in the GRADE assessment process.
Results
Description of studies
Results of the search
We ran searches for the original review in December 2002 and identified three included trials (McPhail 1989; Nezu 1991; Willner 2002). We repeated the searches in February 2007 with the addition of a new search term 'anger management'. We identified one new included study (Taylor 2005).
We updated the searches in May 2013 and April 2014, having revised the strategy to include search terms to describe behavioural phenotypes associated with aggression. Two additional studies met the inclusion criteria (Singh 2013; Willner 2013).
In August 2020, we updated the searches and found 3450 records. After removal of duplicates, we screened 2772 records, of which we considered 59 new titles and abstracts to be relevant. After obtaining the full papers, six additional studies met the inclusion criteria, giving a total of 12 included studies in this review (Hassiotis 2009; Hassiotis 2018; Ingersoll 2017; Jones 2021; Singh 2016; Singh 2020a). We also assessed a study from the previous version of the review that had been awaiting classification, and included it (Collado‐Castillo 2010).
In March 2022, we ran a top‐up search to include trials that had been conducted since August 2020. We found two additional eligible studies (Hall 2020; Kostulski 2021), giving a total of 15 included studies in this review (Figure 1). Two additional studies that were not identified by our search were recommended for screening by an external peer reviewer but were not eligible for inclusion.
Our top‐up search also revealed the protocol for a study that would be eligible (Farris 2020), but had not published its results and therefore is included in section for ongoing studies.
Included studies
Participants
The total number of participants from the 15 included studies was 921. The number of participants in each study ranged from 12 (McPhail 1989) to 245 (Hassiotis 2018). Two studies had a majority of females (Jones 2021; McPhail 1989), one study included only men (Taylor 2005), and the remaining studies had a majority of men. One study enrolled children between the ages of 3 and 10 years (Hall 2020), one study enrolled children and adolescents between the ages of 6 and 18 years (Kostulski 2021), and one study enrolled adolescents and young people between the ages of 12 and 20 years (Ingersoll 2017). Four studies specified a minimum age of 18 years in their inclusion criteria (Hassiotis 2009; Hassiotis 2018; Jones 2021; Taylor 2005), and one study specified an age range of 20 to 65 years as part of their inclusion criteria (Collado‐Castillo 2010). Ethnicity was only available for four studies with all reporting a majority of white participants (Hassiotis 2009; Hassiotis 2018; Nezu 1991; Taylor 2005).
Most included studies focused on participants with mild or mild‐to‐moderate IDs. Taylor 2005 included participants with mild‐to‐borderline IQs. Hassiotis 2009 and Hassiotis 2018 included participants with severe and profound IDs. Singh 2016 focused on participants with severe and profound IDs only. McPhail 1989 included participants across the range of IDs.
Hall 2020 focused on participants with Fragile X syndrome.
Location
Six studies were conducted in the UK, seven studies were in the US, one study was conducted in Canada and one study was conducted in Germany. Fourteen studies recruited participants from the community and one study focused exclusively on an inpatient forensic service (Taylor 2005).
Study design
Fourteen studies were RCTs, of which two were cluster‐RCTs (Hassiotis 2018; Willner 2013). One study was a quasi‐randomised trial (Willner 2002). One study randomly assigned participants to three groups (Collado‐Castillo 2010), and the remainder were trials comprising two groups.
Interventions
Type of intervention
The studies varied in their interventions. Five studies included behavioural interventions in the form of PBS (Hassiotis 2009; Hassiotis 2018), RIT (Ingersoll 2017), and parent training (tele‐health‐enabled behavioural training (FCT) (Hall 2020) and parent management training (PMT) (Kostulski 2021)). Six studies included CBT‐based interventions: five studies of anger management (Collado‐Castillo 2010; Nezu 1991; Taylor 2005; Willner 2002; Willner 2013), one study examined DBT (Jones 2021), one study of mindfulness (Singh 2013), and one used modified relaxation training (McPhail 1989). Two studies combined behavioural and CBT‐based approaches (MBPBS) (Singh 2016; Singh 2020a).
Six studies delivered a manualised intervention (Hassiotis 2018; Jones 2021; Nezu 1991; Singh 2013; Taylor 2005; Willner 2013). One study reported using a 'scripted curriculum' (Collado‐Castillo 2010).
Delivery of intervention
In seven studies, carers or support staff ('lay therapists') delivered therapy after receiving training on the intervention (Hall 2020; Hassiotis 2018; Kostulski 2021; Singh 2013; Singh 2016; Singh 2020a; Willner 2013). In Willner 2013, psychologists provided fortnightly supervision to the 'lay therapists' delivering group‐based CBT. In Ingersoll 2017, teacher‐support staff provided the intervention. In the other studies, trained therapists/psychologists or the researchers delivered the therapy. In one study, the therapists received peer supervision (Taylor 2005).
Adherence to intervention
Ten studies performed fidelity checks and adherence to the intervention (Hall 2020; Hassiotis 2018; Ingersoll 2017; Jones 2021; Nezu 1991; Singh 2013; Singh 2016; Singh 2020a; Taylor 2005; Willner 2013).
Type of comparator
Seven studies compared the intervention to a no treatment control or wait‐list control group (Jones 2021; Kostulski 2021; Nezu 1991; Singh 2013; Taylor 2005; Willner 2002; Willner 2013). One study compared two interventions with a wait‐list control (three arms; Collado‐Castillo 2010). Three studies compared the intervention to TAU (Hassiotis 2009; Hassiotis 2018; Hall 2020). Two studies compared the intervention with PBS (Singh 2016; Singh 2020a) and two studies used an 'active' control arm, which involved another activity that was not an intervention (Ingersoll 2017; McPhail 1989).
Outcomes and measures
See Appendix 3 for the list of scales and their associated abbreviations used in this review.
Primary outcome
Change in aggressive behaviour (frequency/severity of incidents)
Four studies reported the number of incidents of aggressive behaviour. Singh 2016 recorded the number of aggressive events per week, the use of physical restraints per week, weekly use of STAT medication (i.e. requiring immediate use), and use of one‐to‐one staffing per week. Singh 2020a recorded the same number of variables but did not stipulate over what timeframe they recorded these events. Singh 2013 recorded the number of incidents of physical and verbal aggression on a scale developed for the study, which parents and support staff rated. McPhail 1989 recorded the level of disruptive behaviour (aggressive, verbal, 'movement' and 'other') observed by staff three times a week over 45 minutes. The scale was also developed for the study.
Four studies used the Aberrant Behaviour Checklist (ABC) (Aman 1985) to measure the severity of aggressive behaviour (Hall 2020; Hassiotis 2009; Hassiotis 2018; Ingersoll 2017). Willner 2013 measured the severity/intensity of aggressive (challenging) behaviour using the Hyperactivity and Irritability subscales of the ABC; Modified Overt Aggression Scale (MOAS); and the Controllability Beliefs Scale (CBS). Kostulski 2021 used the German version of the Developmental Behaviour Checklist (DBC), the Verhaltensfragebebogen für Kinder mit Entwicklungsstörungen (VFE; Einfeld 2007). Staff/carers reported all these measures. Collado‐Castillo 2010 used the short form of the Aggression Questionnaire (AQ), which both staff and the participant reported.
Change in ability to control anger
The most frequently used measure was the PI (Novaco 2003), used by Jones 2021, Taylor 2005, Willner 2002, and Willner 2013 to record self‐reported levels of aggression in anger‐inducing situations. Jones 2021, Willner 2002, and Willner 2013 also used the PI to record key worker and carer‐reported levels of aggression in anger‐inducing situations.
Willner 2013 used the Profile of Anger Coping Skills (PACS; Willner 2005) to measure self‐reported ability to control anger, which key workers and carers reported.
Taylor 2005 measured anger disposition and capacity to regulate anger using the Novaco Anger Scale (NAS) and the Anger Expression (AX) subscale from the Spielberger State‐Trait Anger Expression Inventory (STAXI; Spielberger 1996). Taylor 2005 also used the Ward Anger Rating Scale (WARS; Novaco 1994), which a member of staff completed to investigate anger attributes.
Nezu 1991 used the Role‐Play Test of Anger Arousing Situations (RPT). Other measures used were the Anger Inventory (AI; Benson 1992; Willner 2002) and the PACS (Willner 2013). Collado‐Castillo 2010 used the Social Information Processing Skills (SIPS) Interview.
Due to the relatively large number of measures, for the purpose of this review, we only reported the results from the PI and PACS. Several studies used the PI and Willner 2013, which was a large study with low risk of bias, used both measures.
Change in adaptive functioning
Ingersoll 2017 used the Matson Evaluation of Social Skills for Individuals with Severe Mental Retardation (MESSIER; Matson 1998) to assess social skills, the Unstructured Imitation Assessment (UIA) and the Motor Imitation Scale (MIS) to assess spontaneous imitation of actions in an unstructured and structured environment.
Nezu 1991 used the Adaptive Behaviour Scale – Revised‐Part II (ABS‐R‐II) as a measure of adaptive functioning and used the Problem‐Solving Task (PST) to measure problem‐solving ability in hypothetical situations.
Secondary outcomes
Change in mental state
Hassiotis 2018 screened for mental health and autism spectrum disorders using the Mini PAS‐ADD and used this to discern between common mental disorder (CMD) and severe mental illness (SMI). Jones 2021 screened for mental health problems using the Reiss Screen for Maladaptive Behaviour (REISS; Reiss 1988).
Two studies included psychological symptoms. Nezu 1991 used the Brief Symptom Inventory (BSI) and the Subjective Units of Distress Scale (SUDS) to measure psychological distress. Willner 2013 measured depression using the Glasgow Depression Scale – Learning Disability (GDS‐LD); anxiety using the Glasgow Anxiety Scale – Learning Disability (GAS‐LD) and self‐esteem using the Rosenberg Self Esteem Scale (RSES).
Change in quality of life
One study assessed quality of life. Willner 2013 assessed quality of life using the COMQoL‐ID (Cummins 1997).
Frequency of service utilisation (and costs)
Hassiotis 2018 and Willner 2013 measured costs of service utilisation using the CSRI (Beecham 1992). Hassiotis 2018 included the CSRI but there was insufficient information to include in our analysis. However in their subgroup analysis of participants, these data were available and included. Two studies provided gross costs in their studies, but were unable to provide additional information required for the analysis (Singh 2016; Singh 2020a).
Duration of follow‐up
Eight studies provided follow‐up data. Hassiotis 2009; McPhail 1989; Nezu 1991; Singh 2013; Willner 2002 had follow‐up at three months; Taylor 2005 had follow‐up at four months; Hassiotis 2009 and Hassiotis 2018 had follow‐up at six months; Willner 2013 had follow‐up at ten months and Hassiotis 2018 had follow‐up at 12 months. There were no follow‐up data for Collado‐Castillo 2010; Hall 2020; Ingersoll 2017; Jones 2021; Kostulski 2021; Singh 2016; and Singh 2020a.
Sources of funding
The South Essex Partnership University Foundation NHS Trust funded Hassiotis 2009. The Autism Speaks Pilot Treatment Award funded Ingersoll 2017. The Department of Psychiatry Research Grant Fund, Queen's University, Canada funded Jones 2021. The National Institute of Health Research (NIHR), Health Technology Assessment Programme funded Hassiotis 2018 and Willner 2013. There were no funding statements for McPhail 1989; Nezu 1991; Singh 2013; Singh 2016; Singh 2020a; Taylor 2005; and Willner 2002.
Excluded studies
In this update, we excluded 48 studies after examining the full text following both searches and two additional studies that were recommended for screening by a reviewer. Of these, one was a systematic review; 31 did not meet the criteria for eligibility of participants (e.g. the minimum number/ proportion of participants with IDs was less than 70%); nine did not report specifically about outwardly aggressive behaviour and nine did not have an eligible intervention.
We reported a subset of key excluded studies that would plausibly be expected to be included in the review but were not, and the main reason for their exclusion in the Characteristics of excluded studies table.
In the previous version of the review, 32 studies were excluded (Ali 2015). One study, Hassiotis 2009, was previously excluded but has been included in this review following closer inspection and consensus within the review team that the study meets the inclusion criteria.
Studies awaiting classification
There are no studies awaiting classification.
Ongoing studies
One study was ongoing and has been added to the Characteristics of ongoing studies table (Farris 2020).
Risk of bias in included studies
See risk of bias tables beneath the Characteristics of included studies table, risk of bias summary (review authors' judgements about the risk of bias in the included studies Figure 2) and the risk of bias graph (review authors' judgements about each risk of bias item presented as percentages across all included studies; Figure 3).
Risk of bias summary: review authors' judgements about each risk of bias item for each included study.
Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies.
Allocation
Random sequence generation
Eleven studies were at low risk of selection bias for sequence generation (Hall 2020 (online software tool); Hassiotis 2009 (computer‐driven randomisation based on a block size of four); Hassiotis 2018 (an independent web‐based randomisation system using permuted block randomisation); Ingersoll 2017 (randomly assigned by a coin flip); Jones 2021 (randomised in blocks); Kostulski 2021 (shuffled deck of cards); Nezu 1991 (randomisation by coin tossing; Nezu 2000 [pers comm]); Singh 2016 (a random number generator); Singh 2020a (a random number generator); Taylor 2005 (computer‐generated randomisation codes); Willner 2013 (minimisation with a random component set at 80%)). One study was at high risk of bias (Willner 2002 (randomisation by alternate allocation of referrals)). Three studies were at unclear risk of bias as the methods were not reported (Collado‐Castillo 2010; McPhail 1989; Singh 2013).
Allocation concealment
Seven studies were at low risk of selection bias for allocation (Hall 2020 (study used a web‐based system with permutated blocks); Hassiotis 2009 (an independent administrator held a set of sealed envelopes); Hassiotis 2018 (used a central web‐based system); Jones 2021 (the principal investigator who was not involved in the recruitment of participants conducted randomisation); Kostulski 2021 (used a deck of shuffled cards making it difficult to predict which group participants would be randomised into); Taylor 2005 (the groups were balanced on key variables, but the person involved in this did not conduct any of the assessments, and other personnel involved in the study were not aware of group allocation); Willner 2013 (randomisation was conducted by a centralised system and all the centres and participants were assessed prior to randomisation)). One study was at high risk of selection bias (Willner 2002 (reported that they did not conceal allocation of their participants)). Seven studies made no reference to whether any process of concealment of allocation was used and were at unclear risk of bias (Collado‐Castillo 2010; Ingersoll 2017; McPhail 1989; Nezu 1991; Singh 2013; Singh 2016; Singh 2020a).
Blinding
In all studies, blinding of participants was not possible due to the nature of the intervention, and we thus rated all studies at high risk of performance bias.
Six studies were at low risk of detection bias as outcome assessors were blinded to the allocation group (Hassiotis 2009; Hassiotis 2018; Jones 2021; McPhail 1989; Nezu 1991; Willner 2013). However, Willner 2013 reported that the allocation group often became apparent during conversations with participants and carers. Seven studies were at high risk of detection bias as outcome assessors were not blinded to group allocation (Collado‐Castillo 2010; Hall 2020; Singh 2013; Singh 2016; Singh 2020a; Taylor 2005; Willner 2002). Ingersoll 2017 reported that the blinding of the assessors was possible for only two of the outcomes and therefore this study was at unclear risk of bias. Kostulski 2021 did not provide information on who conducted the assessments and was at unclear risk of bias.
Incomplete outcome data
Ten studies were at low risk of attrition bias (Collado‐Castillo 2010; Hall 2020; Hassiotis 2009; Hassiotis 2018; McPhail 1989; Singh 2013; Singh 2016; Willner 2002; Willner 2013). There were no dropouts in the study by Collado‐Castillo 2010 and Singh 2013. Singh 2016 reported that two participants did not receive the intervention due to medical reasons and were not included in the analysis. Overall dropout rates were low, and therefore this study has been rated as low risk of bias. Jones reported a dropout rate of 36% for the intervention arm and 20% for the control arm but an intention‐to‐treat analysis was conducted using the last observation carried forward. Although Willner 2013 had many dropouts (19% data missing), there was the same number of dropouts in both arms (17 in each), and they conducted an intention‐to‐treat analysis. Hassiotis 2009 had a low attrition rate where one participant died and there was one dropout in the intervention group and one died in the control group, but data were analysed based on last observation carried forward. McPhail 1989 excluded two participants from the analysis (one from each group) despite completing treatment, and Willner 2002 excluded one participant from the analysis and one participant dropped out, but we rated both of these studies at low risk, as the dropouts or removal of participants from the analysis were unrelated to the outcome.
One study was at high risk of attrition bias because the four participants who dropped out were in the treatment group and there was no intention‐to‐treat analysis (Taylor 2005). In previous versions of this review (see Hassiotis 2004; Hassiotis 2008), we attempted to contact the authors for clarification and to obtain the missing data, but received no response. No further attempts were made to contact the authors in this updated version. Therefore, subsequent analyses are based on the data provided in the papers.
Four studies were at unclear risk of bias (Ingersoll 2017; Kostulski 2021Nezu 1991; Singh 2020a). Kostulski 2021 reported seven families dropping out after randomisation and two families dropping out in both arms subsequently. Ingersoll 2017 reported that one participant dropped out from the control arm but missing data were only reported for two outcome measures and it was unclear if there were missing data for the other outcomes. Nezu 1991 did not report whether there were any dropouts. Singh 2020a reported that one participant had dropped out of the intervention arm and six had dropped out of the comparison arm but did not provide reasons for the dropouts.
Selective reporting
Thirteen studies were at low risk of reporting bias as the prespecified outcomes appeared to have been reported (Collado‐Castillo 2010; Hall 2020; Hassiotis 2009; Hassiotis 2018; Ingersoll 2017; Jones 2021; Nezu 1991; Singh 2013; Singh 2016; Singh 2020a; Taylor 2005; Willner 2002; Willner 2013). One study was at high risk of reporting bias, as follow‐up data were not included in the analysis (McPhail 1989). One study was at unclear risk of reporting bias, as there was no mention of whether an intention‐to‐treat analysis was performed (Kostulski 2021).
Other potential sources of bias
Three studies were at low risk of other bias (Hall 2020 (characteristics of participants across both arms were equivocal with a 95.7% fidelity to treatment); Hassiotis 2009 (no other risk of bias, including differences in baseline characteristics of the participants); McPhail 1989 (no other risk of bias, including differences in baseline characteristics of the participants)). Two studies were at high risk of other bias Collado‐Castillo 2010 (only published as a PhD thesis and not as a peer‐reviewed paper); Willner 2002 (used non‐validated instruments)). Ten studies were at unclear risk of other bias. Taylor 2005 and Nezu 1991 included participants with mental health problems who were also receiving psychotropic medication, and it was unclear whether inclusion of these participants influenced the results. Kostulski 2021, Singh 2013, Singh 2016, Singh 2020a, and Willner 2013 had a strong allegiance to the intervention and their desire for the intervention to succeed could have affected data collection, but the likelihood of this impacting the results was unclear. Ingersoll 2017 did not have a formal write‐up of a study protocol, and it was unclear when follow‐up occurred. Hassiotis 2018 acknowledged the limitations were low treatment fidelity and reach of the intervention. Jones 2021 acknowledged difficulties in trying to adhere to a randomised controlled design.
Effects of interventions
See: Summary of findings 1 Summary of findings table ‐ Anger management compared to wait‐list control for outwardly aggressive behaviour in people with intellectual disability; Summary of findings 2 Summary of findings table ‐ Positive behavioural support compared to treatment as usual for people with outwardly aggressive behaviour and intellectual disability; Summary of findings 3 Summary of findings table ‐ Meditation based on mindfulness compared to treatment as usual for people with outwardly aggressive behaviour and intellectual disability; Summary of findings 4 Summary of findings table ‐ Mindfulness based on positive behaviour support compared to positive behaviour support in outwardly aggressive behaviour in people with intellectual disability
Comparison 1: anger management compared to wait‐list control
Three studies examined the effectiveness of anger management based on CBT compared to wait‐list control (Taylor 2005; Willner 2002; Willner 2013). One study used assertiveness and problem‐solving and has been included here as it is based on similar principles to anger management (Nezu 1991). One study compared two CBT‐based interventions for anger management (problem‐solving and cognitive re‐appraisal) with wait‐list control (Collado‐Castillo 2010). See summary of findings Table 1.
Primary outcomes
Change in aggressive behaviour (frequency/severity of incidents)
Willner 2013 reported the adjusted MD between treatment and control groups for their measures but, for the purpose of this review, we reported the unadjusted MD and corresponding 95% CIs.
We found moderate‐certainty evidence that anger management may reduce hyperactivity and irritability reported by key‐workers 16 weeks post‐treatment compared to wait‐list control (ABC – Hyperactivity subscale: MD −4.80, 95% CI −7.60 to −2.00; P < 0.001; 1 study, 159 participants; Analysis 1.1); ABC – Irritability subscale: MD −3.50, 95% CI −6.21 to −0.79; P = 0.01; 1 study, 158 participants; Analysis 1.2). Evidence was downgraded one level due to imprecision (results based on one study). These differences were not maintained at 10 months' follow‐up (ABC – Hyperactivity subscale: MD −1.20, 95% CI −3.98 to 1.58; P = 0.40; 1 study, 150 participants; low‐certainty evidence; Analysis 1.1; ABC – Irritability subscale: MD 0.80, 95% CI −1.89 to 3.49; P = 0.56; 1 study, 150 participants; low‐certainty evidence; Analysis 1.2). Evidence was downgraded two levels due to imprecision (results from one study and the CIs crossed the null effect). Home carers reported no reduction in the ABC – Hyperactivity subscale at 16 weeks or 10 months (16 weeks: MD −3.00, 95% CI −7.17 to 1.17; P = 0.16; 1 study, 104 participants; low‐certainty evidence; 10 months: MD −2.40, 95% CI −7.20 to 2.40; P = 0.33; 1 study, 84 participants; low‐certainty evidence; Analysis 1.3). There were also no changes on the ABC Irritability subscale reported by home carers at 16 weeks or 10 months (16 weeks: MD −3.10, 95% CI −7.10 to 0.90; P = 0.13; 1 study, 104 participants; low‐certainty evidence; 10 months: MD −2.20, 95% CI −6.90 to 2.50; P = 0.36; 1 study, 84 participants; low‐certainty evidence; Analysis 1.4). Evidence was downgraded two levels due to imprecision (results from one study and the CIs crossed the null effect).
Anger management did not reduce aggressive behaviour when measured using the MOAS and reported by key workers at 16 weeks' post‐treatment and 10 months' follow‐up (16 weeks: MD 1.70, 95% CI −2.55 to 5.95; P = 0.43; 1 study, 158 participants; low‐certainty evidence; 10 months: MD 0.40, 95% CI −3.62 to 4.42; P = 0.85; 1 study, 140 participants; low‐certainty evidence; Analysis 1.5). The evidence was downgraded two levels due to imprecision (results from one study and the CIs crossed the null effect). Similarly, there were no changes in the MOAS scores reported by carers at 16 weeks and 10 months (16 weeks: MD −4.00, 95% CI −10.55 to 2.55; P = 0.23; 1 study, 103 participants; very low‐certainty evidence; 10 months: MD −0.50, 95% CI −6.91 to 5.91; P = 0.88; 1 study, 83 participants; very low‐certainty evidence; Analysis 1.6). The evidence was downgraded three levels due to imprecision (results from one study and wide CIs that crossed the null effect).
Collado‐Castillo 2010 (24 participants) compared two CBT‐based interventions for anger management (problem‐solving and cognitive re‐appraisal) with wait‐list control. They examined aggressive behaviour using the AQ, which has five subscales. We found very low‐certainty evidence that problem‐solving might not reduce aggressive behaviour post‐treatment, compared to the control group, as reported by staff and participants using the Physical Aggression subscale (staff: MD −1.00, 95% CI −2.52 to 0.52; P = 0.20; 1 study, 24 participants; Analysis 1.7; participants: MD −1.00, 95% CI −2.64 to 0.64; P = 0.23; Analysis 1.7); Verbal Aggression subscale (staff: MD 0.08, 95% CI −1.84 to 2.00; P = 0.94; Analysis 1.8; participants: MD 0.00, 95% CI −1.25 to 1.25; P = 1.00; Analysis 1.8); Anger subscale (staff: MD 0.08, 95% CI −1.19 to 1.35; P = 0.90; Analysis 1.9; participants: MD 0.08, 95% CI −1.20 to 1.36; P = 0.90; Analysis 1.9) and Hostility subscale (staff: MD 0.75, 95% CI −0.68 to 2.18; P = 0.30; Analysis 1.10; participants: MD −1.00, 95% CI −84.18 to 82.18; P = 0.98; Analysis 1.10). We found very low‐certainty evidence that the intervention might reduce aggression on the Indirect Aggression subscale as reported by participants but not by staff (staff: MD −0.58, 95% CI −2.20 to 1.04; P = 0.48; Analysis 1.11; participants: MD −2.00, 95% CI −3.33 to −0.67; P = 0.003; Analysis 1.11).
We found very low‐certainty evidence from Collado‐Castillo 2010 (24 participants) that cognitive re‐appraisal might not reduce aggressive behaviour post‐treatment, compared to the control group, as reported by staff and participants using the Physical Aggression subscale (staff: MD −1.42, 95% CI −2.92 to 0.08; P = 0.06; Analysis 1.12; participants: MD −0.08, 95% CI −1.86 to 1.70; P = 0.93; Analysis 1.12); Verbal subscale (staff: MD 0.75, 95% CI −1.04 to 2.54; P = 0.41; Analysis 1.13; participants: MD 0.17, 95% CI −0.81 to 1.15; P = 0.74; Analysis 1.13); Anger subscale (staff: MD 0.50, 95% CI −1.04 to 2.04; P = 0.52; Analysis 1.14; participants: MD 0, 95% CI −1.21 to 1.21; P = 1.00; Analysis 1.14); Hostility subscale (staff: MD 0.58, 95% CI −0.82 to 1.98; P = 0.42; Analysis 1.15; participants: MD 0.34, 95% CI −1.06 to 1.74; P = 0.63; Analysis 1.15) and Indirect Aggression subscale (staff: MD −0.58, 95% CI −2.14 to 0.98; P = 0.47; Analysis 1.16; participants: MD −0.17, 95% CI −1.46 to 1.12; P = 0.80; Analysis 1.16). The evidence was downgraded three levels due to high risk of bias (due to study limitations arising from lack of blinding of outcome assessors and several unclear risk of bias assessments) and imprecision (results from one study and CIs crossed the null effect).
Change in ability to control anger
Provocation Inventory
Three studies examined participant's ability to control anger, measured using the participant self‐reported PI (Taylor 2005; Willner 2002; Willner 2013). When the results were combined in a meta‐analysis, we found very low‐certainty evidence that anger management may improve ability to control anger compared to the control group post‐treatment (MD −8.38, 95% CI −14.05 to −2.71; P = 0.004; 3 studies, 212 participants; Analysis 1.17). There was no evidence of heterogeneity (Chi2 = 2.04, P= 0.36; I2 = 2%). The evidence was downgraded three levels due to high risk of bias because of study limitations (lack of blinding of outcome assessors in two studies and in one study there was high risk of bias from random sequence generation and allocation concealment) and imprecision due to wide CIs and the CIs in two studies crossed the null effect).
There was very low‐certainty evidence from Taylor 2005 (36 participants) that anger management might improve scores post‐treatment on the PI – Irritations subscale (MD −2.36, 95% CI −4.67 to −0.05; P = 0.05; Analysis 1.18). The evidence was downgraded due to high risk of bias (downgraded two levels due to study limitations arising from lack of blinding of outcome assessors and incomplete reporting of outcome data) and imprecision (evidence based on one study with small sample size). There was very low‐certainty evidence from Taylor 2005 (36 participants) that anger management does not improve scores on the Disrespect subscale (MD −1.16, 95% CI −3.30 to 0.98; P = 0.29; Analysis 1.19), Unfairness subscale (MD −2.04, 95% CI −4.45 to 0.37; P = 0.10; Analysis 1.20), Annoying Traits subscale (MD −0.85, 95% CI −4.00 to 2.30; P = 0.60; Analysis 1.21), and Frustration subscale (MD −0.80, 95% CI −3.95 to 2.35; P = 0.62; Analysis 1.22). The evidence was downgraded three levels due to high risk of bias (study limitations arising from lack of blinding of outcome assessors and incomplete reporting of outcome data) and imprecision (results from one study with small sample size and CIs crossed the null effect).
There was moderate‐certainty evidence that anger management may improve ability to control anger using the PI post‐treatment, as reported by key workers and carers (MD −5.76, 95% CI −9.70 to −1.82; P = 0.004; 2 studies, 3 groups, 279 participants; Analysis 1.23) (Willner 2002; Willner 2013). There was no evidence of heterogeneity (Chi2 = 3.83, P =0.15, I2 = 48%). The evidence was downgraded one level due to imprecision (wide CIs).
There was very low‐certainty evidence from Taylor 2005 (36 participants) that anger management might not improve ability to control anger on the PI as reported by participants at four months (MD −4.96, 95% CI −15.82 to 5.90; P = 0.37; Analysis 1.17) and on all the subscales of the PI (Irritations subscale: MD −1.41, 95% CI −3.83 to 1.01; P = 0.25; Analysis 1.18; Disrespect subscale: MD −0.30, 95% CI −2.82 to 2.22; P = 0.82; Analysis 1.19; Unfairness subscale: MD −0.19, 95% CI −2.53 to 2.15; P = 0.87; Analysis 1.20; Annoying Traits subscale: MD −1.55, 95% CI −4.24 to 1.14; P = 0.26; Analysis 1.21; Frustration subscale: MD −1.50, 95% CI −4.20 to 1.20; P = 0.28; Analysis 1.22). The evidence was downgraded three levels due to high risk of bias due to study limitations (lack of blinding of outcome assessors and incomplete reporting of outcome data) and imprecision (results from one study with small sample size and wide CIs).
There was low‐certainty evidence from Willner 2013 (143 participants) that anger management probably does not improve ability to control anger at 10 months (MD −3.70, 95% CI −10.55 to 3.15; P = 0.29; Analysis 1.17). Evidence was downgraded two levels due to imprecision (results from one study with small sample size and wide CIs).
There was low‐certainty evidence from Willner 2013 (2 groups, 245 participants) that anger management probably does not lead to improvements in ability to control anger as reported by both key workers and carers at 10 months' follow‐up (MD −4.58, 95% CI −9.27 to 0.11; P = 0.06; Analysis 1.23). Evidence was downgraded two levels due to inconsistency (heterogeneity (Chi 2 = 4.81; P = 0.03; I2 = 79%)) and imprecision (the CIs crossed the null effect).
Profile of Anger Coping Skills
There was low‐certainty evidence that anger management may improve participants' ability to manage anger on the self‐reported PACS post‐treatment compared to the control group (MD 8.70, 95% CI 0.98 to 16.42; P = 0.03; 1 study, 156 participants; Analysis 1.24). The evidence was downgraded two levels due to imprecision (results from one study and wide CIs). There was very low‐certainty evidence that anger management might not improve ability to control anger on the self‐reported PACS at 10 months' follow‐up (MD 7.70, 95% CI −0.72 to 16.12; P = 0.07; 1 study, 138 participants; Analysis 1.24). The evidence was downgraded three levels due to imprecision (results from one study and wide CIs that crossed the null effect).
There was moderate‐certainty evidence that anger management may improve ability to control anger using the PACS as reported by key workers and carers at post‐treatment and at 10 months' follow‐up when their results were combined (post‐treatment: MD 6.15, 95% CI 2.08 to 10.21; P = 0.003; I2 = 0%; 1 study, 2 groups, 260 participants; 10 months' follow‐up: MD 5.40, 95% CI 0.72 to 10.09; P = 0.02; I2 = 47%; 1 study, 2 groups, 225 participants; Analysis 1.25). The evidence was downgraded one level due to imprecision (wide CIs).
Change in adaptive functioning
There was very low‐certainty evidence that anger management may improve adaptive functioning compared to control measured using the ABS‐R‐II (MD −21.73, 95% CI −36.44 to −7.02; P = 0.004; 1 study, 28 participants; Analysis 1.26). The evidence was downgraded three levels due risk of bias (unclear risk of allocation concealment and incomplete outcome data) and imprecision (results from one study with small sample size and wide CIs).
Adverse effects
None of the studies in this comparison reported data on adverse effects.
Secondary outcomes
Change in mental state
There was very low‐certainty evidence that anger management may improve psychiatric symptoms (reduction of score on BSI: MD −0.48, 95% CI −0.79 to −0.17; P = 0.002; 1 study, 28 participants; Analysis 1.27; Nezu 1991) and lower levels of subjective distress measured using the SUDS in the intervention group compared to the control group post‐treatment (MD −4.36, 95% CI −6.85 to −1.87; P < 0.001; 1 study, 28 participants; Analysis 1.28). The evidence was downgraded three levels due to risk of bias (from unclear risk of allocation concealment and incomplete outcome data) and imprecision (results from one study with small sample size and wide CIs).
There was low‐certainty evidence that anger management probably does not improve symptoms of depression, anxiety and self‐esteem at 16 weeks' post‐treatment (depression: MD −0.70, 95% CI −3.06 to 1.66; P = 0.56; 1 study, 157 participants; Analysis 1.29; anxiety: MD −2.30, 95% CI −5.39 to 0.79; P = 0.15; 1 study, 154 participants; Analysis 1.30; self‐esteem: MD 0.30, 95% CI −0.98 to 1.58; P = 0.65; 1 study, 141 participants; Analysis 1.31) (Willner 2013). At 10 months' follow‐up, there was low‐certainty evidence that anger management probably does not improve symptoms of depression, anxiety or self‐esteem (depression: MD 0.20, 95% CI −2.15 to 2.55; P = 0.87; 1 study, 144 participants; Analysis 1.29; anxiety: MD 0.40, 95% CI −2.59 to 3.39; P = 0.79; 1 study, 143 participants; Analysis 1.30; self‐esteem: MD −0.70, 95% CI −2.22 to 0.82; P = 0.37; 1 study, 134 participants; Analysis 1.31). Evidence was downgraded two levels due to imprecision (results from one study and the CIs crossed the null effect).
Change in quality of life
There was very low‐certainty evidence from one study that anger management probably does not improve quality of life (Willner 2013), measured using the Comprehensive Quality of Life Scale, compared to the control group at post‐treatment or 10 months' follow‐up (post‐treatment: MD −5.60, 95% CI −18.11 to 6.91; P = 0.38; 1 study, 129 participants; Analysis 1.32; 10 months' follow‐up: MD −0.60, 95% CI −13.26 to 12.06; P = 0.93; 1 study, 140 participants; Analysis 1.32). The evidence was downgraded three levels due to imprecision (results from one study and wide CIs that crossed the null effect).
Frequency of service utilisation (and costs)
There was very low‐certainty evidence that anger management probably does not reduce service utilisation and costs compared to control at 10 months' follow‐up (MD GBP 102.99, 95% CI −117.16 to 323.14; P = 0.36; 1 study, 133 participants; Analysis 1.33). The evidence was downgraded three levels due to imprecision (results from one study and wide CIs that crossed the null effect).
Other secondary outcomes
None of the studies in this comparison reported data on change in (additional) medication, change in care needs or user satisfaction (or dissatisfaction).
Comparison 2: positive behavioural support or behavioural support team compared to treatment as usual
Two studies compared PBS to TAU (Hassiotis 2009; Hassiotis 2018). See summary of findings Table 2.
Primary outcomes
Change in aggressive behaviour (frequency/severity of incidents)
Neither Hassiotis 2009 nor Hassiotis 2018 demonstrated a reduction in behaviour at six months post‐treatment, measured using the ABC Total score. However, when we combined the results in a meta‐analysis, at six months post‐treatment, we found moderate‐certainty evidence that PBS may reduce severity of aggressive behaviour compared to TAU (MD −7.78, 95% CI −15.23 to −0.32; P = 0.04; 2 studies, 275 participants; Analysis 2.1). The evidence was downgraded one level due to imprecision (wide CIs). There was no evidence of heterogeneity (Chi2 = 0.87, P = 0.35; I2 = 0%).
There was low‐certainty evidence that PBS probably does not reduce severity of incidents of aggressive behaviour at 12 months (MD −5.20, 95% CI −13.27 to 2.87; P = 0.21; 1 study, 225 participants; Analysis 2.1; Hassiotis 2018). The evidence was downgraded two levels due to imprecision (results from one study and wide CIs).
Hassiotis 2009 and Hassiotis 2018 also reported scores on the individual subscales of the ABC, including the Irritability subscale, which is most closely associated with aggression. The paper provided data on the median and interquartile ranges and not the means and SDs. Therefore, we presented the analysis that was reported in these papers (analyses not shown). Hassiotis 2009 reported that there were no changes in the Irritability subscale between PBS and TAU (MD −0.21, 95% CI −0.50 to 0.008; P = 0.162). Similarly, Hassiotis 2018 found no differences in the Irritability subscale (MD −0.041, 95% CI −0.22 to 0.14).
Adverse effects
Hassiotis 2018 reported that 29 participants experienced 45 serious adverse events that were unrelated to the study. Of these, 26 serious adverse events were in the intervention arm (including one person dying) and 19 in the TAU arm. Two participants were admitted to a psychiatric hospital for mental illness and challenging behaviour, and the remaining serious adverse events were related to hospital admissions in relation to physical health issues, such as falls, or deterioration in chronic conditions such as epilepsy and respiratory conditions. The number of people experiencing adverse events in each arm was not stated.
Hassiotis 2009 did not make specific reference to serious adverse events but stated that one participant was admitted to a psychiatric hospital and another three were admitted to a general hospital for physical health complaints. The report did not clarify which arm the participants were allocated to.
Other primary outcomes
Neither study in this comparison reported data on change in ability to control anger or change in adaptive functioning.
Secondary outcomes
Change in mental state
Hassiotis 2018 reported the prevalence of mental disorders assessed using the Mini PAS‐ADD. There was low‐certainty evidence that PBS probably does not reduce the prevalence of CMDs at six or 12 months (6 months: OR 1.44, 95% CI 0.83 to 2.49; P = 0.19; 1 study, 214 participants; Analysis 2.2; 12 months: OR 0.95, 95% CI 0.56 to 1.62; P = 0.86, 1 study, 225 participants; Analysis 2.2) or SMI at six or 12 months (6 months: OR 1.66, 95% CI 0.76 to 3.62; P = 0.20; 1 study, 214 participants; Analysis 2.3; 12 months: OR 0.87, 95% CI 0.42 to 1.80; P = 0.71; 1 study, 225 participants; Analysis 2.3). Evidence was downgraded two levels due to imprecision (results from one study and CIs crossed the null effect).
Hassiotis 2009 used the PAS‐ADD checklist to report the prevalence of mental disorders. There was very low‐certainty evidence that PBS might not reduce the prevalence of affective disorders, psychotic disorders or organic mental disorders at six months (affective disorders: OR 0.82, 95% CI 0.24 to 2.81; P = 0.75; 1 study, 60 participants; Analysis 2.4; psychotic disorders: OR 1.56, 95% CI 0.24 to 10.05; P = 0.64; 1 study, 60 participants; Analysis 2.5; organic mental disorders: OR 0.36, 95% CI 0.06 to 2.01; P = 0.24; 1 study, 60 participants; very low certainty evidence; Analysis 2.6). The evidence was downgraded three levels due to imprecision (results from one study and wide CIs that crossed the null effect).
Frequency of service utilisation (and costs)
We found very low‐certainty evidence that PBS probably does not reduce service utilisation and costs at six, 12 or 36 months (6 months: MD GBP −2200.00, 95% CI −5186.22 to 786.23; P = 0.15; 1 study, 61 participants; Analysis 2.7; 12 months: MD GBP −448.00, 95% CI −1660.83 to 764.83; P = 0.47; 1 study, 225 participants; Analysis 2.7; 36 months: MD 81.00, 95% CI −1610.96 to 1772.96; P = 0.93; 1 study, 180 participants; Analysis 2.7; Hassiotis 2018). Evidence was downgraded three levels due to imprecision (results from one study and wide CIs that crossed the null effect).
Other secondary outcomes
Neither study in this comparison reported data on change in (additional) medication, change in care needs, change in quality of life or user satisfaction (or dissatisfaction).
Comparison 3: meditation based on mindfulness compared to wait‐list control
Singh 2013 compared mindfulness to wait‐list control. See summary of findings Table 3.
Primary outcomes
Change in aggressive behaviour (frequency/severity of incidents)
There was very low‐certainty evidence that mindfulness might reduce the number of incidents of both physical and verbal aggression (physical aggression: MD −2.80, 95% CI −4.37 to −1.23; P < 0.001; 1 study; 34 participants; Analysis 3.1; verbal aggression: MD −3.30, 95% CI −5.05 to −1.55; P < 0.001; 1 study, 34 participants; Analysis 3.2). The evidence was downgraded three levels due to risk of bias from study limitations (unclear risk of selection bias and high risk of detection bias due to lack of blinding of study assessors) and imprecision (results from one study with small sample size and wide CIs).
Other primary outcomes
The study in this comparison reported no data on change in ability to control anger, change in adaptive functioning or adverse effects.
Secondary outcomes
The study in this comparison reported no data on our secondary outcomes: change in mental state, change in (additional) medication, change in care needs, change in quality of life, frequency service utilisation (and costs) and user satisfaction (or dissatisfaction).
Comparison 4: carer training in mindfulness‐based positive behaviour support compared to positive behaviour support
Two studies compared carer training in MBPBS to training in PBS only (Singh 2016; Singh 2020a). See summary of findings Table 4.
Primary outcomes
Change in aggressive behaviour (frequency/severity of incidents)
Both studies examined differences in carer reports of aggressive incidents (Singh 2016; Singh 2020a). When we combined the results in a meta‐analysis there was low‐certainty evidence that carer training in MBPBS may reduce reports of aggressive incidents compared to PBS (MD −10.27, 95% CI −14.87 to −5.67; P < 0.001; 128 participants; Analysis 4.1). However, heterogeneity between the studies was high (Chi2 = 7.72, P = 0.005; I2 = 87%). The heterogeneity may be explained by differences in samples (in Singh 2016, participants had severe and profound ID and in Singh 2020a, participants had mild‐to‐moderate ID). The evidence was downgraded two levels due to risk of bias (lack of blinding of outcome assessors) and inconsistency (high levels of heterogeneity).
There was low‐certainty evidence that MBPBS may reduce carer reports of the number of staff injuries and peer injuries post‐treatment compared to PBS alone (staff injuries: MD −1.46, 95% CI −1.95 to −0.97; P < 0.001; 1 study, 80 participants; Analysis 4.2; peer injuries: MD −2.47, 95% CI −2.93 to −2.01; P < 0.001; 1 study, 80 participants; Analysis 4.3; Singh 2020a). The evidence was downgraded two levels due to risk of bias from study limitations (lack of blinding of outcome assessors) and imprecision (small sample size).
Other primary outcomes
Neither study in this comparison reported on change in ability to control anger, change in adaptive functioning or data on adverse effects.
Secondary outcomes
Change in (additional) medication
Both Singh 2016 and Singh 2020a examined the use of emergency medication. When the results were combined in a meta‐analysis, we found very low‐certainty evidence that it might not reduce use of emergency medication in the intervention group post‐treatment (MD −3.98, 95% CI −9.68 to 1.71; P = 0.17; 2 studies, 128 participants; Analysis 4.4). The evidence was downgraded three levels due to risk of bias, imprecision (CIs crossed the null effect) and high levels of heterogeneity (Chi 2 = 50.03; df = 1; I2 = 98%).
Change in care needs
Both Singh 2016 and Singh 2020a investigated carer reports on the use of physical restraints and need for one‐to‐one staffing. When the results were combined in a meta‐analysis, there was very low‐certainty evidence that the intervention may reduce the use of physical restraint in the MBPBS group compared to the PBS group post‐treatment (MD −7.15, 95% CI −13.16 to −1.13; P = 0.02; 2 studies, 128 participants; Analysis 4.5). The results were downgraded three levels due to risk of bias (unclear risk of attrition bias and lack of blinding of assessors) and imprecision (wide CIs and inconsistency due to high levels of heterogeneity (Chi 2 = 59.35, degrees of freedom (df) = 1, I2 = 98%)). There was moderate‐certainty evidence that the intervention may reduce the need for one‐to‐one staffing (MD −3.94, 95% CI −4.33 to −3.54; P < 0.001; 2 studies, 128 participants; Analysis 4.6). Heterogeneity was not significant (Chi 2 = 3.06; df = 1, P = 0.08, I2 = 67%) and, therefore, the evidence was only downgraded one level (risk of bias).
Frequency of service utilisation (and costs)
We were unable to calculate the MD for the staffing costs incurred by the two groups due to insufficient data (SDs were unavailable). Singh 2016 reported that there were savings of over 78% (USD 457,920) and Singh 2020a report a saving of USD 512,418 in the intervention arm compared to the control arm at 40 weeks.
Other secondary outcomes
Neither study in this comparison reported on change in mental state, quality of life or user satisfaction (or dissatisfaction).
Comparison 5: adapted dialectical behaviour therapy compared to no treatment control
One study compared adapted dialectical behaviour therapy (aDBT) to no treatment (Jones 2021).
Primary outcomes
Change in ability to control anger
There was very low‐certainty evidence that adapted dialectal behaviour therapy might not improve client ratings of ability to control anger on the self‐reported or carer‐reported PI (self‐reported: MD 14.60, 95% CI −1.37 to 30.57; P = 0.07; 1 study, 20 participants; Analysis 5.1; carer ratings: MD −6.30, 95% CI −16.50, 3.90; P = 0.23; 1 study, 20 participants; Analysis 5.2). The evidence was downgraded three levels due to imprecision (results from one study with small sample size and wide CIs that crossed the null effect).
Other primary outcomes
The study in this comparison reported no data on change in aggressive behaviour, change in adaptive functioning or adverse effects.
Secondary outcomes
Change in mental state
There was very low‐certainty evidence that aDBT might not improve client ratings of mental health based on the REISS (MD 2.10, 95% CI −5.32 to 9.52; P = 0.58; 1 study, 20 participants; Analysis 5.3). However, there was very low‐certainty evidence of a difference in favour of the control group post‐treatment in REISS carer ratings (MD 10.70, 95% CI 2.53 to 18.87; P = 0.01; 1 study, 20 participants; Analysis 5.4). The evidence was downgraded three levels due to imprecision (results from one study with small sample size and wide CIs).
Other secondary outcomes
The study in this comparison reported no data on change in (additional medication), change in care needs, change in quality of life, frequency of service utilisation (and costs) and user satisfaction (or dissatisfaction).
Comparison 6: reciprocal imitation training compared to an active control group (one‐to‐one instruction on individualised treatment plan)
One study compared RIT to an active control (Ingersoll 2017).
Primary outcomes
Change in aggressive behaviour (frequency/severity of incidents)
There was low‐certainty evidence that RIT may reduce aggressive behaviour measured using the ABC post‐treatment compared to the individualised treatment plan (MD −18.61, 95% CI −36.46 to −0.76; P = 0.04; 1 study, 19 participants; Analysis 6.1). The evidence was downgraded three levels due to imprecision (results from one study with small sample size and wide CIs).
Change in adaptive functioning
There was low‐certainty evidence that RIT may improve adaptive functioning post‐treatment measured using the MESSIER (MD 19.69, 95% CI 2.48 to 36.90; P = 0.02; 1 study, 19 participants; Analysis 6.2). The evidence was downgraded two levels due to imprecision (very small sample size and very wide CIs).
However, there was very low‐certainty evidence that the intervention might not improve adaptive functioning measured using the UIA and MIS (UIA: MD 14.72, 95% CI −10.92 to 40.36; P = 0.26; 1 study, 19 participants; Analysis 6.3; MIS: MD 17.59, 95% CI −13.07 to 48.25; P = 0.26; 1 study, 19 participants; Analysis 6.4). The evidence was downgraded three levels due to study limitations arising from unclear risk of bias in several domains (e.g. allocation concealment, blinding of outcome assessors and missing data) and due to imprecision (results from one study with very small sample size and very wide CIs that crossed the null effect).
Other primary outcomes
The study in this comparison reported no data on change in ability to control anger and adverse effects.
Secondary outcomes
The study in this comparison reported no data on our secondary outcomes: change of mental state, change in (additional) medication, change in care needs, change in quality of life, frequency of service utilisation (and costs) and user satisfaction (or dissatisfaction).
Comparison 7: parenting interventions using behavioural approaches compared to treatment as usual or wait‐list control
Two studies compared parents received training on the use of behavioural approaches in managing their child's problem behaviours to TAU or wait‐list control (Hall 2020; Kostulski 2021).
Primary outcomes
Change in aggressive behaviour (frequency/severity of incidents)
There were insufficient data from Hall 2020 to enable the calculation of the MD and 95% CI using the ABC‐C – Irritability subscale post‐treatment. Therefore, we provided a narrative description of the results reported in the paper. In the intervention group (FCT), the mean score on the irritability subscale of the ABC‐C decreased by 6.44 points from the baseline score at 12 weeks (42.6% reduction). In the TAU group, there was a decrease of 1.35 points from the baseline score at 12 weeks (9.13% reduction). When comparing the two groups, the intervention group showed greater improvement in scores (P < 0.001, Cohen's d = 0.65).
Hall 2020 reported data on the proportion of participants who had a positive response (at least 25% improvement in ABC – Irritability subscale scores). There was very low‐certainty evidence that FCT may increase the proportion of participants with a positive response (25% improvement in ABC – Irritability subscale scores) compared to TAU (OR 4.86, 95% CI 1.43 to 16.50; P = 0.01; 1 study, 48 participants; very low certainty evidence; Analysis 7.1). The study was downgraded three levels because of imprecision (results from one study and wide CIs) and due to risk of bias from study limitations (high risk of detection bias due to lack of blinding of study assessors).
In Kostulski 2021, there was insufficient data to enable the calculation of the MD or OR for the disruptive/antisocial subscale of the German Version of the DBC. Only the total scores are provided. Therefore, only a narrative description is provided here. The authors reported medium effect sizes for the reduction in disruptive behaviour/antisocial behaviour in the group receiving the intervention, although the results were not significant (F = 13.53, P = 0.59, Cohen's d = 0.66).
Adverse effects
Parents receiving FCT completed the Treatment Acceptability Rating Form – Revised (TARF‐R) (Hall 2020). The scores on the Side Effects subscale of the TARF‐R were reported as being low at baseline (mean 7.8, SD 2.9) and remained low throughout treatment, indicating that there were minimal adverse effects reported by the intervention group.
Other primary outcomes
Neither study in this comparison reported data on change in ability to control anger or change in adaptive functioning.
Change in mental state
There were medium effects reported on the subscales of the DBC related to anxiety in favour of the intervention group, but the results were not significant (F = 3.61, P = 0.288, Cohen's d = 0.059).
User satisfaction data (or dissatisfaction)
Parents receiving FCT measured satisfaction using the TARF‐R (Hall 2020). The authors reported that there were high scores at baseline on the Reasonableness subscale (mean 17.9, SD 2.8); Effectiveness subscale (mean 16.9, SD 2.8) and Willingness subscale (mean 18.4, SD 1.9), which remained consistently high during the treatment, indicating high levels of satisfaction. The scores on the Costs subscale (mean 12.4, SD 2.6) and Disruptiveness subscale (mean 12.4, SD 2.4) were in the middle range.
Other secondary outcomes
Neither study in this comparison reported data on change in (additional) medication, change in care needs, change in quality of life or frequency of service utilisation (and costs).
Comparison 8: modified relaxation training compared to an active control group (story reading)
One study compared modified relaxation training versus an active control group (story reading) (McPhail 1989).
Primary outcomes
Change in aggressive behaviour (frequency/severity of incidents)
McPhail 1989 did not report the SDs for outcome measures in the treatment and control groups, and, therefore, we could not calculate CIs. Only a narrative description is provided here. Overall, there was a 74% reduction in total incidents of disruptive behaviours in the treatment group from a mean of 14.3 incidents per person per 30 minutes at baseline to 3.6 incidents per person per 30 minutes on the final day of treatment. The control group showed no improvement (mean number of incidents at baseline 10.2 and at end‐of‐treatment 10.9; MD 7.3 number of incidents (F(6, 48) = 2.3, P = 0.04)). The mean measure of verbal disruptive behaviour decreased from 5.6 incidents per person per 30 minutes to 2.5 incidents per person per 30 minutes at the end of treatment, whereas the number of incidents in the control group increased from 3.6 to 5.4 (groups × weeks effect F(3,24) = 3.3, P = 0.03; MD 2.9). There was also an improvement in aggressive disruptive behaviour from 1.6 incidents at baseline in the treatment group to no incidents at the end of treatment whereas the number of incidents in the control group increased from 1.9 to 4 (group effect F(1,3) = 10.2, P = 0.04; MD 4.0). There were no differences between the two groups in the mean number of movement disruptive incidents and 'other' disruptive incidents. However, at three‐month follow‐up, disruption levels had returned to those at baseline.
Other primary outcomes
The study in this comparison reported no data on change in ability to control anger, change in adaptive functioning or adverse effects.
Secondary outcomes
The study in this comparison reported no data on change in mental state, change in (additional) medication, change in care needs, change in quality of life, frequency of service utilisation (and costs), and user satisfaction data (or dissatisfaction).
Discussion
Summary of main results
This updated review contains nine new studies (Collado‐Castillo 2010; Hall 2020; Hassiotis 2009; Hassiotis 2018; Ingersoll 2017; Jones 2021; Kostulski 2021; Singh 2016; Singh 2020a), making a total of 15 included studies that examined a range of behavioural (PBS, RIT and parent training) and CBT approaches (individual‐ and group‐based anger management, relaxation, mindfulness‐based meditation, MBPBS and aDBT) for outwardly aggressive behaviour in adults with IDs. Six studies were conducted in the UK (Hassiotis 2009; Hassiotis 2018; McPhail 1989; Taylor 2005; Willner 2002; Willner 2013); seven were conducted in the US (Collado‐Castillo 2010; Hall 2020; Ingersoll 2017; Nezu 1991; Singh 2013; Singh 2016; Singh 2020a), one in Canada (Jones 2021), and one in Germany (Kostulski 2021). Participants were recruited from community‐based facilities in all but one of the included studies. In general, the samples in these studies map onto the typical characteristics of people with disruptive behaviours likely to be seen by community‐based services in the UK. The one study that did not recruit from community‐based facilities included detained male service users with serious aggression (Taylor 2005). The participants in the review had borderline/mild‐to‐severe/profound IDs, although most had mild‐to‐moderate ID.
Only three studies made reference to adverse events (Hall 2020; Hassiotis 2009; Hassiotis 2018), and two studies had long‐term data beyond six months (Hassiotis 2018; Willner 2013).
Summary of the evidence for each comparison
Five studies examined anger management (Collado‐Castillo 2010; Nezu 1991; Taylor 2005; Willner 2002; Willner 2013). There was moderate‐certainty evidence that anger management is likely to improve severity of aggressive behaviour measured using the Irritability and Hyperactivity subscales of the ABC (Willner 2013). Very low‐certainty evidence from three studies suggest that anger management might improve ability to control anger post‐treatment when measured using the PI reported by participants but not at four or 10 months' follow‐up (Taylor 2005; Willner 2002; Willner 2013). There was low‐certainty evidence that it may improve ability to control anger measured using the PACS (Willner 2013). There was moderate‐certainty evidence that anger management may improve ability to control anger post‐treatment and at 10 months using the PACS reported by carers. There was very low‐certainty evidence that anger management may improve adaptive functioning and psychiatric symptoms and very low certainty evidence that it probably does not improve quality of life or reduce service utilisation and costs (Willner 2013).
Two studies investigated PBS (Hassiotis 2009; Hassiotis 2018). When the studies were combined in a meta‐analysis, there was moderate‐certainty evidence that PBS may reduce aggressive behaviour post‐treatment. There was low‐certainty evidence that PBS probably does not improve aggressive behaviour at 12 months, or improve mental state. There was very low‐certainty evidence that PBS might not reduce service utilisation or costs at both six and 12 months.
Three studies examined mindfulness‐based interventions (Singh 2013; Singh 2016; Singh 2020a). When mindfulness was compared to wait‐list control, there was very low‐certainty evidence that it may reduce incidents of verbal and physical aggression (Singh 2013). Two studies examined carer training on MBPBS, which compared the intervention with PBS. When the results were combined in a meta‐analysis, there was low‐certainty evidence that MBPBS may reduce incidents of aggression, staff injuries and peer injuries; moderate‐certainty evidence that it may reduce one‐to‐one staffing and very low‐certainty evidence that it may reduce physical restraints.
One study explored the use of aDBT (Jones 2021). There was very low‐certainty evidence that aDBT might not improve the ability to control anger and mental state compared to a no treatment control group. Another study compared RIT to an active control arm (Ingersoll 2017). We found low‐certainty evidence that it may reduce aggressive behaviour measured using the ABC and that it may improve social adaptive functioning. One study examined a parenting intervention (FCT) for children with Fragile X syndrome. We found very low‐certainty evidence that the parenting intervention might increase the proportion of participants who had a positive response (25% improvement in ABC – Irritability subscale scores) compared to TAU.
Based on all the available data, there was moderate‐ to very low‐certainty evidence that cognitive and behavioural interventions may be effective in reducing aggressive behaviour, improving ability to control anger and adaptive functioning, and in reducing psychopathology and medication use in the short term. However, there is a paucity of long‐term follow‐up data in most studies included in this review and of studies that are of high quality. Therefore, there is a need for further research using multicentre clinical trials to elucidate the effectiveness of psychological and social interventions for aggressive behaviour in pragmatic settings beyond ascertaining efficacy.
Overall completeness and applicability of evidence
Participants
The studies in this review included participants with varying degrees of ID from mild to severe. While most of the studies were of adults, three studies were of children or adolescents (one in young children (Hall 2020), and two in children and adolescents (Ingersoll 2017; Kostulski 2021). There was only one study that included participants who specifically had a genetic phenotype (Fragile X syndrome; Hall 2020).
Most studies included participants in the community, except for Taylor 2005, which included a forensic inpatient sample. This may limit the generalisability of the findings to individuals living in the community, who have less severe aggressive behaviour, compared to those in inpatient or more restrictive settings.
There is a lack of evidence from low‐ and middle‐income countries and, therefore, the results cannot be generalised to countries where the provision of services for people with IDs may be very limited.
Interventions
This updated review includes interventions such as PBS, aDBT, RIT and parent training, which were not included in the previous versions of the review. However, there were only single studies of DBT and RT, with these studies having very small sample sizes. The evidence for these particular interventions remains limited and further studies are warranted.
Comparisons
This updated review includes studies that have compared "enhanced PBS" with PBS alone. Singh 2016 and Singh 2020a compared MBPBS with PBS alone. This might reflect changes in clinical practice. For example, in the UK, PBS has become standard care and there maybe ethical challenges in conducting clinical trials using a 'no treatment' or wait‐list control group. One study also had an 'active' control arm where participants received the same number of sessions (Ingersoll 2017).
Outcomes
Almost all the comparisons included a measure of aggressive behaviour, which was assessed either using a validated measure (e.g. ABC) or was based on the number of recorded incidents. The use of the ABC does have some limitations in that it is an indirect measure of aggressive behaviour. Some studies used the ABC to measure general changes in challenging behaviours rather than specifically aggressive behaviour. The ABC includes items that measure other types of behavioural problems and, therefore, changes in the total ABC score may not reflect changes in aggressive behaviour. This also applies (but to a lesser degree) to the Irritability subscale of the ABC. However, it can be argued that irritability and emotional dysregulation are on a continuum with aggressive behaviour (Melville 2016; Stringaris 2018), and may precede acts of physical aggression, which provides support for the ABC (especially the Irritability subscale) as an appropriate measure of outwardly aggressive behaviour.
The review provides accumulating evidence for the potential role of behavioural and cognitive behavioural approaches in reducing aggressive behaviours, albeit the evidence varies from moderate to very low certainty depending on the intervention. There is less evidence for other outcomes that were relevant to this review. Only two comparisons mentioned ability to control anger (anger management and DBT); two comparisons mentioned adaptive behaviour (anger management and RIT); two comparison mentioned mental state (anger management and PBS); one comparison mentioned quality of life (anger management) and two comparisons mentioned service use and costs (anger management and PBS). Only two studies examining MBPBS provided evidence in relation to the use of additional medication and care needs. Therefore, there is a clear need for more RCTs that examine these outcomes. However, this updated review did include a wider range of outcomes compared to the previous version of the review.
Quality of the evidence
We used the GRADE approach to assess the certainty of the body of evidence. The certainty of evidence was very low for most outcomes included in the review (see summary of findings Table 1; summary of findings Table 2; summary of findings Table 3; summary of findings Table 4). However, there was moderate‐certainty evidence that anger management and PBS may reduce the severity of aggressive behaviour (both downgraded one level due to imprecision (anger management: results from one study; PBS: wide CIs)).
The main reasons for downgrading the certainty of evidence across all outcomes were due to risk of bias (particularly selection and performance bias); imprecision (single studies with small sample sizes, wide CIs and CIs crossed the null effect) and inconsistency (statistical heterogeneity).
In relation to the methodological quality and risk of bias in studies, one study did not perform random sequence generation and concealment of allocation (Willner 2002); almost all studies were unable to blind participants and personnel; there was lack of blinding of outcome assessors in five studies (Singh 2013; Singh 2016; Singh 2020a; Taylor 2005; Willner 2002); there was selective reporting in one study (McPhail 1989), and one study had other biases (use of non‐validated measures, Willner 2002).
Potential biases in the review process
The search terms and strategies that we used in the review were comprehensive, and we are confident that we identified all relevant studies. Nonetheless, there is a possibility that studies may have been missed. The review process could be further enhanced by more extensive handsearching of journals and contact with experts in the field. We ensured that members of the review team who were not authors of studies that were being reviewed, conducted the screening and made decisions about the inclusion of these studies. This mitigated against potential bias that may have been introduced if authors of this review were able to influence the inclusion and reporting of studies in which they were involved.
Agreements and disagreements with other studies or reviews
The findings from this review are broadly similar to the previous version of this review (Ali 2015). Eight of the nine new studies provided further evidence in support of the use of behavioural and cognitive‐behavioural approaches in improving some participant outcomes. However, no studies found significant improvements in quality of life or reduction of costs.
The results are consistent with other systematic reviews and meta‐analysis of behavioural and cognitive‐behavioural interventions in the management of aggressive behaviour. One systematic review found large effect sizes for the treatment of anger using CBT, but they included studies that were not randomised and studies that did not include a control group (Nicoll 2013). Another meta‐analysis found that behavioural interventions were effective in reducing different types of challenging behaviour. However, the included studies were single‐case studies (Didden 2006). Chapman 2013 found that mindfulness improved aggression and led to reduced use of physical restraint. Two reviews found a positive effect of treatment on any type for problem behaviours (Heyvaert 2010; Heyvaert 2012). However, the treatments included pharmacological as well as psychological therapies in managing a range of challenging behaviours, and small case studies. Another systematic review of psychological therapies indicated that individual CBT may be effective for treating depression in people with IDs (Vereenooghe 2013), whilst interventions aiming at improving interpersonal relationships were least effective. The National Institute for Health and Care Excellence (NICE) guidelines on challenging behaviour and learning disabilities examined a range of psychosocial and pharmacological interventions across all types of challenging behaviours (NICE 2015). Five studies using CBT and behavioural approaches were included in the analysis and there was evidence for a reduction in the severity of the target behaviour at the end of the intervention, but the evidence was rated as low or very low according to the GRADE criteria.
Study flow diagram.
Risk of bias summary: review authors' judgements about each risk of bias item for each included study.
Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies.
Comparison 1: Anger management vs wait‐list control, Outcome 1: Aggressive behaviour: severity of incidents: Aberrant Behaviour Checklist (ABC) – Hyperactivity subscale – key worker report
Comparison 1: Anger management vs wait‐list control, Outcome 2: Aggressive behaviour: severity of incidents: ABC – Irritability subscale – key worker report
Comparison 1: Anger management vs wait‐list control, Outcome 3: Aggressive behaviour: severity of incidents: ABC – Hyperactivity subscale – home carer report
Comparison 1: Anger management vs wait‐list control, Outcome 4: Aggressive behaviour: severity of incidents: ABC – Irritability subscale – home carer report
Comparison 1: Anger management vs wait‐list control, Outcome 5: Aggressive behaviour: severity of incidents: Modified Overt Aggression Scale (MOAS) – key worker report
Comparison 1: Anger management vs wait‐list control, Outcome 6: Aggressive behaviour: severity of incidents: MOAS – home carer report
Comparison 1: Anger management vs wait‐list control, Outcome 7: Aggressive behaviour: severity of incidents: Aggression Questionnaire – Physical Aggression subscale (problem‐solving intervention vs control)
Comparison 1: Anger management vs wait‐list control, Outcome 8: Aggressive behaviour: severity of incidents: Aggression Questionnaire – Verbal Aggression subscale (problem‐solving intervention vs control)
Comparison 1: Anger management vs wait‐list control, Outcome 9: Aggressive behaviour: severity of incidents: Aggression Questionnaire – Anger subscale (problem‐solving intervention vs control)
Comparison 1: Anger management vs wait‐list control, Outcome 10: Aggressive behaviour: severity of incidents: Aggression Questionnaire – Hostility subscale (problem‐solving intervention vs control)
Comparison 1: Anger management vs wait‐list control, Outcome 11: Aggressive behaviour: severity of incidents: Aggression Questionnaire – Indirect Aggression subscale (problem‐solving vs control)
Comparison 1: Anger management vs wait‐list control, Outcome 12: Aggressive behaviour: severity of incidents: Aggression Questionnaire – Physical Aggression subscale (cognitive re‐appraisal intervention vs control)
Comparison 1: Anger management vs wait‐list control, Outcome 13: Aggressive behaviour: severity of incidents: Aggression Questionnaire – Verbal Aggression subscale (cognitive re‐appraisal vs control)
Comparison 1: Anger management vs wait‐list control, Outcome 14: Aggressive behaviour: severity of incidents: Aggression Questionnaire – Anger subscale (cognitive re‐appraisal vs control)
Comparison 1: Anger management vs wait‐list control, Outcome 15: Aggressive behaviour: severity of incidents: Aggression Questionnaire – Hostility subscale (cognitive re‐appraisal vs control)
Comparison 1: Anger management vs wait‐list control, Outcome 16: Aggressive behaviour: severity of incidents: Aggression Questionnaire – Indirect Aggression subscale (cognitive re‐appraisal vs control)
Comparison 1: Anger management vs wait‐list control, Outcome 17: Ability to control anger: Provocation Inventory (PI) – self‐report
Comparison 1: Anger management vs wait‐list control, Outcome 18: Ability to control anger: PI – Irritations subscale
Comparison 1: Anger management vs wait‐list control, Outcome 19: Ability to control anger: PI – Disrespect subscale
Comparison 1: Anger management vs wait‐list control, Outcome 20: Ability to control anger: PI – Unfairness subscale
Comparison 1: Anger management vs wait‐list control, Outcome 21: Ability to control anger: PI – Annoying Traits subscale
Comparison 1: Anger management vs wait‐list control, Outcome 22: Ability to control anger: PI – Frustrations subscale
Comparison 1: Anger management vs wait‐list control, Outcome 23: Ability to control anger: PI – key worker and carer reports
Comparison 1: Anger management vs wait‐list control, Outcome 24: Ability to control anger: Profile of Anger Coping Skills (PACS) – self‐report
Comparison 1: Anger management vs wait‐list control, Outcome 25: Ability to control anger: PACS – key worker and carer reports
Comparison 1: Anger management vs wait‐list control, Outcome 26: Adaptive functioning: Adaptive Behaviour Scale – Revised, Part II (ABS‐II)
Comparison 1: Anger management vs wait‐list control, Outcome 27: Mental state: psychiatric symptoms measured using the Brief Symptom Inventory (BSI)
Comparison 1: Anger management vs wait‐list control, Outcome 28: Mental state: psychological distress measured using the Subjective Units of Distress Scale (SUDS)
Comparison 1: Anger management vs wait‐list control, Outcome 29: Mental state: depression: Glasgow Depression Scale – Intellectual Disability (GDS‐ID)
Comparison 1: Anger management vs wait‐list control, Outcome 30: Mental state: anxiety: Glasgow Anxiety Scale – Intellectual Disability (GAS‐ID)
Comparison 1: Anger management vs wait‐list control, Outcome 31: Self‐esteem: Rosenberg Self‐Esteem Scale (SES)
Comparison 1: Anger management vs wait‐list control, Outcome 32: Quality of life: Comprehensive Quality of Life Scale: Intellectual Disability (COMQoL‐ID)
Comparison 1: Anger management vs wait‐list control, Outcome 33: Costs of service utilisation: Client Service Receipt Inventory (CSRI): cost per person per week of health and social care resource (in British pounds)
Comparison 2: Positive behavioural support (PBS) or behavioural support team vs treatment as usual (TAU), Outcome 1: Aggressive behaviour: Aberrant Behaviour Checklist (ABC) Total
Comparison 2: Positive behavioural support (PBS) or behavioural support team vs treatment as usual (TAU), Outcome 2: Mental state: Common Mental Disorder (CMD) assessed using the Mini‐Psychiatric Assessment Schedule for Adults with Developmental Disabilities (Mini PAS‐ADD)
Comparison 2: Positive behavioural support (PBS) or behavioural support team vs treatment as usual (TAU), Outcome 3: Mental state: severe mental illness (SMI) assessed using the Mini PAS‐ADD
Comparison 2: Positive behavioural support (PBS) or behavioural support team vs treatment as usual (TAU), Outcome 4: Mental state: prevalence of affective disorders assessed using the PAS‐ADD checklist
Comparison 2: Positive behavioural support (PBS) or behavioural support team vs treatment as usual (TAU), Outcome 5: Mental state: prevalence of psychotic disorders assessed by the PAS‐ADD checklist
Comparison 2: Positive behavioural support (PBS) or behavioural support team vs treatment as usual (TAU), Outcome 6: Mental state: prevalence of organic mental disorder assessed by the PAS‐ADD Checklist
Comparison 2: Positive behavioural support (PBS) or behavioural support team vs treatment as usual (TAU), Outcome 7: Costs of service utilisation: Client Service receipt Inventory (CSRI) – Total mean health and social care costs per participant
Comparison 3: Meditation based on mindfulness vs wait‐list control, Outcome 1: Aggressive behaviour: frequency of incidents: number of incidents of physical aggression per week during treatment (12 weeks)
Comparison 3: Meditation based on mindfulness vs wait‐list control, Outcome 2: Aggressive behaviour: frequency of incidents: number of incidents of verbal aggression per week during treatment (12 weeks)
Comparison 4: Carer training in mindfulness‐based positive behaviour support (MBPBS) vs positive behaviour support (PBS), Outcome 1: Aggressive behaviour: aggressive events – carer reports
Comparison 4: Carer training in mindfulness‐based positive behaviour support (MBPBS) vs positive behaviour support (PBS), Outcome 2: Aggressive behaviour: staff injuries – carer reports
Comparison 4: Carer training in mindfulness‐based positive behaviour support (MBPBS) vs positive behaviour support (PBS), Outcome 3: Aggressive behaviour: peer injuries – carer reports
Comparison 4: Carer training in mindfulness‐based positive behaviour support (MBPBS) vs positive behaviour support (PBS), Outcome 4: Reduction in (additional) medication: emergency medication – carer reports
Comparison 4: Carer training in mindfulness‐based positive behaviour support (MBPBS) vs positive behaviour support (PBS), Outcome 5: Reduction in care needs: physical restraints – carer reports
Comparison 4: Carer training in mindfulness‐based positive behaviour support (MBPBS) vs positive behaviour support (PBS), Outcome 6: Reduction in care needs: 1:1 staffing – carer reports
Comparison 5: Adapted dialectical behaviour therapy (aDBT) vs no treatment control, Outcome 1: Ability to control anger: Novaco Anger Scale: Provocation Inventory (NAS:PI) – client ratings
Comparison 5: Adapted dialectical behaviour therapy (aDBT) vs no treatment control, Outcome 2: Ability to control anger: NAS:PI – carer ratings
Comparison 5: Adapted dialectical behaviour therapy (aDBT) vs no treatment control, Outcome 3: Mental state: Reiss Screen for Maladaptive Behaviour (REISS) – client ratings
Comparison 5: Adapted dialectical behaviour therapy (aDBT) vs no treatment control, Outcome 4: Mental state: REISS – carer ratings
Comparison 6: Reciprocal imitation training (RIT) vs 1:1 instruction on individualised treatment plan, Outcome 1: Aggressive behaviour: Aberrant Behaviour Checklist (ABC) – Residential
Comparison 6: Reciprocal imitation training (RIT) vs 1:1 instruction on individualised treatment plan, Outcome 2: Adaptive functioning: Matson Evaluation of Social Skills for Individuals with Severe Mental Retardation (MESSIER)
Comparison 6: Reciprocal imitation training (RIT) vs 1:1 instruction on individualised treatment plan, Outcome 3: Adaptive functioning: Unstructured Imitation Assessment (UIA)
Comparison 6: Reciprocal imitation training (RIT) vs 1:1 instruction on individualised treatment plan, Outcome 4: Adaptive functioning: Motor Imitation Scale (MIS)
Comparison 7: Parent training in behavioural approaches vs treatment as usual (TAU) or wait‐list control, Outcome 1: Proportion of participants who achieved a positive response on the ABC (≥ 25% reduction in scores)
| Anger management compared to wait‐list control for outwardly aggressive behaviour in people with intellectual disability | ||||||
| Patient or population: outwardly aggressive behaviour in people with intellectual disability | ||||||
| Outcomes | Anticipated absolute effects* (95% CI) | Relative effect | № of participants | Certainty of the evidence | Comments | |
|---|---|---|---|---|---|---|
| Risk with wait‐list control | Risk with anger management | |||||
| Aggressive behaviour: severity of incidents: Aberrant Behaviour Checklist (ABC – irritability subscale – key worker report – post‐treatment (Analysis 1.2.1) | The mean aggressive behaviour: severity of incidents: Aberrant Behaviour Checklist (ABC – irritability subscale – key worker report – post‐treatment (Analysis 1.2.1) was 11.0 | MD 3.5 lower | ‐ | 158 | ⊕⊕⊕⊝ | — |
| Ability to control anger: Provocation Inventory (PI) – self‐report – post‐treatment (Analysis 1.17.1) | The mean ability to control anger: Provocation Inventory (PI) – self‐report – post‐treatment (Analysis 1.17.1) was 54.85 | MD 8.38 lower | ‐ | 212 | ⊕⊝⊝⊝ | — |
| Adaptive functioning: Adaptive Behaviour Scale – Revised, Part II (ABS‐II) (Analysis 1.26) | The mean adaptive functioning: Adaptive Behaviour Scale – Revised, Part II (ABS‐II) (Analysis 1.26) was 74.9 | MD 21.73 lower | ‐ | 28 | ⊕⊝⊝⊝ | — |
| Mental state: psychiatric symptoms measured using the Brief Symptom Inventory (BSI) (Analysis 1.27) | The mean mental state: psychiatric symptoms measured using the Brief Symptom Inventory (BSI) (Analysis 1.27) was 1.22 | MD 0.48 lower | ‐ | 28 | ⊕⊝⊝⊝ | — |
| Quality of life: Comprehensive Quality of Life Scale: Intellectual Disability (COMQoL‐ID) – post‐treatment (Analysis 1.32.1) | The mean quality of life: Comprehensive Quality of Life Scale: Intellectual Disability (COMQoL‐ID) – post‐treatment (Analysis 1.32.1) was 99.9 | MD 5.6 lower | ‐ | 129 | ⊕⊝⊝⊝ | — |
| Costs of service utilisation: Client Service Receipt Inventory (CSRI): cost per person per week of health and social care resource (in British pounds (GBP)) – follow‐up: 10 months (Analysis 1.33) | The mean costs of service utilisation: Client Service Receipt Inventory (CSRI): cost per person per week of health and social care resource (in British pounds (GBP)) – follow‐up: 10 months (Analysis 1.33) was GBP 867.09 | MD 102.99 higher | ‐ | 133 | ⊕⊝⊝⊝ | — |
| *The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). | ||||||
| GRADE Working Group grades of evidence | ||||||
| See interactive version of this table: https://gdt.gradepro.org/presentations/#/isof/isof_question_revman_web_430801803348621284. | ||||||
| a Downgraded one level due to imprecision as the results were based on one study. | ||||||
| Positive behavioural support compared to treatment as usual for people with outwardly aggressive behaviour and intellectual disability | ||||||
| Patient or population: people with outwardly aggressive behaviour and intellectual disability | ||||||
| Outcomes | Anticipated absolute effects* (95% CI) | Relative effect | № of participants | Certainty of the evidence | Comments | |
|---|---|---|---|---|---|---|
| Risk with treatment as usual | Risk with positive behavioural support | |||||
| Aggressive behaviour: Aberrant Behaviour Checklist (ABC) Total – at 6 months (Analysis 2.1.1) | The mean aggressive behaviour: Aberrant Behaviour Checklist (ABC) Total – at 6 months (Analysis 2.1.1) was 56.53 | MD 7.78 lower | ‐ | 275 | ⊕⊕⊕⊝ | — |
| Ability to control anger ‐ not measured | ‐ | ‐ | ‐ | ‐ | ‐ | — |
| Adaptive functioning ‐ not measured | ‐ | ‐ | ‐ | ‐ | ‐ | — |
| Mental state: Common Mental Disorder (CMD) – at 6 months (Analysis 2.2.1) (CMD) | 368 per 1000 | 456 per 1000 | OR 1.44 | 215 | ⊕⊕⊝⊝ | — |
| Quality of life ‐ not measured | ‐ | ‐ | ‐ | ‐ | ‐ | — |
| Costs of service utilisation: mean health and social care costs per participant – at 12 months (Analysis 2.7.2) (Cost of service utilisation) | The mean costs of service utilisation: mean health and social care costs per participant – at 12 months (Analysis 2.7.2) was GBP 4051 | MD 448 lower | ‐ | 225 | ⊕⊝⊝⊝ | — |
| *The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). | ||||||
| GRADE Working Group grades of evidence | ||||||
| See interactive version of this table: https://gdt.gradepro.org/presentations/#/isof/isof_question_revman_web_430802235954903680. | ||||||
| a Downgraded one level due to imprecision as confidence intervals were wide. | ||||||
| Meditation based on mindfulness compared to treatment as usual for people with outwardly aggressive behaviour and intellectual disability | ||||||
| Patient or population: people with outwardly aggressive behaviour and intellectual disability | ||||||
| Outcomes | Anticipated absolute effects* (95% CI) | Relative effect | № of participants | Certainty of the evidence | Comments | |
|---|---|---|---|---|---|---|
| Risk with treatment as usual | Risk with meditation based on mindfulness | |||||
| Aggressive behaviour: frequency of incidents: number of incidents of physical aggression per week during treatment (12 weeks) – post‐treatment (Analysis 3.1) | The mean aggressive behaviour: frequency of incidents: number of incidents of physical aggression per week during treatment (12 weeks) – post‐treatment (Analysis 3.1) was 5.80 | MD 2.8 lower | ‐ | 34 | ⊕⊝⊝⊝ | — |
| Ability to control anger ‐ not measured | ‐ | ‐ | ‐ | ‐ | ‐ | No studies were available |
| Adaptive functioning ‐ not measured | ‐ | ‐ | ‐ | ‐ | ‐ | No studies were available |
| Mental state ‐ not measured | ‐ | ‐ | ‐ | ‐ | ‐ | No studies were available |
| Quality of life ‐ not measured | ‐ | ‐ | ‐ | ‐ | ‐ | No studies were available |
| Cost of service utilisation ‐ not measured | ‐ | ‐ | ‐ | ‐ | ‐ | No studies were available |
| *The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). | ||||||
| GRADE Working Group grades of evidence | ||||||
| See interactive version of this table: https://gdt.gradepro.org/presentations/#/isof/isof_question_revman_web_430895877541611244. | ||||||
| a The evidence was downgraded three levels due to risk of bias from study limitations (unclear risk of selection bias and high risk of detection bias due to lack of blinding of study assessors) and imprecision (results from one study with small sample size and wide confidence intervals). | ||||||
| Mindfulness based on positive behaviour support compared to positive behaviour support in outwardly aggressive behaviour in people with intellectual disability | ||||||
| Patient or population: outwardly aggressive behaviour in people with intellectual disability | ||||||
| Outcomes | Anticipated absolute effects* (95% CI) | Relative effect | № of participants | Certainty of the evidence | Comments | |
|---|---|---|---|---|---|---|
| Risk with positive behaviour support | Risk with mindfulness based on positive behaviour support | |||||
| Aggressive behaviour: aggressive events – carer reports – post‐training (Analysis 4.1) | The mean aggressive behaviour: aggressive events – carer reports – post‐training (Analysis 4.1) was 15.78 | MD 10.27 lower | ‐ | 128 | ⊕⊕⊝⊝ | — |
| Ability to control anger ‐ not measured | ‐ | ‐ | ‐ | ‐ | ‐ | — |
| Adaptive functioning ‐ not measured | ‐ | ‐ | ‐ | ‐ | ‐ | — |
| Mental state ‐ not measured | ‐ | ‐ | ‐ | ‐ | ‐ | — |
| Quality of life ‐ not measured | ‐ | ‐ | ‐ | ‐ | ‐ | — |
| Cost of service utilisation ‐ not measured | ‐ | ‐ | ‐ | ‐ | ‐ | — |
| *The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). | ||||||
| GRADE Working Group grades of evidence | ||||||
| See interactive version of this table: https://gdt.gradepro.org/presentations/#/isof/isof_question_revman_web_430892291844998974. | ||||||
| a Downgraded two levels due to risk of bias arising from study limitations (lack of blinding of outcome assessors) and inconsistency arising from heterogeneity. | ||||||
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1.1 Aggressive behaviour: severity of incidents: Aberrant Behaviour Checklist (ABC) – Hyperactivity subscale – key worker report Show forest plot | 1 | Mean Difference (IV, Fixed, 95% CI) | Subtotals only | |
| 1.1.1 Post‐treatment | 1 | 159 | Mean Difference (IV, Fixed, 95% CI) | ‐4.80 [‐7.60, ‐2.00] |
| 1.1.2 At 10‐month follow‐up | 1 | 150 | Mean Difference (IV, Fixed, 95% CI) | ‐1.20 [‐3.98, 1.58] |
| 1.2 Aggressive behaviour: severity of incidents: ABC – Irritability subscale – key worker report Show forest plot | 1 | Mean Difference (IV, Fixed, 95% CI) | Subtotals only | |
| 1.2.1 Post‐treatment | 1 | 158 | Mean Difference (IV, Fixed, 95% CI) | ‐3.50 [‐6.21, ‐0.79] |
| 1.2.2 At 10‐month follow‐up | 1 | 150 | Mean Difference (IV, Fixed, 95% CI) | 0.80 [‐1.89, 3.49] |
| 1.3 Aggressive behaviour: severity of incidents: ABC – Hyperactivity subscale – home carer report Show forest plot | 1 | Mean Difference (IV, Fixed, 95% CI) | Subtotals only | |
| 1.3.1 Post‐treatment | 1 | 104 | Mean Difference (IV, Fixed, 95% CI) | ‐3.00 [‐7.17, 1.17] |
| 1.3.2 At 10‐month follow‐up | 1 | 84 | Mean Difference (IV, Fixed, 95% CI) | ‐2.40 [‐7.20, 2.40] |
| 1.4 Aggressive behaviour: severity of incidents: ABC – Irritability subscale – home carer report Show forest plot | 1 | Mean Difference (IV, Fixed, 95% CI) | Subtotals only | |
| 1.4.1 Post‐treatment | 1 | 104 | Mean Difference (IV, Fixed, 95% CI) | ‐3.10 [‐7.10, 0.90] |
| 1.4.2 At 10‐month follow‐up | 1 | 84 | Mean Difference (IV, Fixed, 95% CI) | ‐2.20 [‐6.90, 2.50] |
| 1.5 Aggressive behaviour: severity of incidents: Modified Overt Aggression Scale (MOAS) – key worker report Show forest plot | 1 | Mean Difference (IV, Fixed, 95% CI) | Subtotals only | |
| 1.5.1 Post‐treatment | 1 | 158 | Mean Difference (IV, Fixed, 95% CI) | 1.70 [‐2.55, 5.95] |
| 1.5.2 At 10‐month follow‐up | 1 | 140 | Mean Difference (IV, Fixed, 95% CI) | 0.40 [‐3.62, 4.42] |
| 1.6 Aggressive behaviour: severity of incidents: MOAS – home carer report Show forest plot | 1 | Mean Difference (IV, Fixed, 95% CI) | Subtotals only | |
| 1.6.1 Post‐treatment | 1 | 103 | Mean Difference (IV, Fixed, 95% CI) | ‐4.00 [‐10.55, 2.55] |
| 1.6.2 At 10‐month follow‐up | 1 | 83 | Mean Difference (IV, Fixed, 95% CI) | ‐0.50 [‐6.91, 5.91] |
| 1.7 Aggressive behaviour: severity of incidents: Aggression Questionnaire – Physical Aggression subscale (problem‐solving intervention vs control) Show forest plot | 1 | Mean Difference (IV, Fixed, 95% CI) | Subtotals only | |
| 1.7.1 Staff report | 1 | 24 | Mean Difference (IV, Fixed, 95% CI) | ‐1.00 [‐2.52, 0.52] |
| 1.7.2 Participant report | 1 | 24 | Mean Difference (IV, Fixed, 95% CI) | ‐1.00 [‐2.64, 0.64] |
| 1.8 Aggressive behaviour: severity of incidents: Aggression Questionnaire – Verbal Aggression subscale (problem‐solving intervention vs control) Show forest plot | 1 | Mean Difference (IV, Fixed, 95% CI) | Subtotals only | |
| 1.8.1 Staff report | 1 | 24 | Mean Difference (IV, Fixed, 95% CI) | 0.08 [‐1.84, 2.00] |
| 1.8.2 Participant report | 1 | 24 | Mean Difference (IV, Fixed, 95% CI) | 0.00 [‐1.25, 1.25] |
| 1.9 Aggressive behaviour: severity of incidents: Aggression Questionnaire – Anger subscale (problem‐solving intervention vs control) Show forest plot | 1 | Mean Difference (IV, Fixed, 95% CI) | Subtotals only | |
| 1.9.1 Staff report | 1 | 24 | Mean Difference (IV, Fixed, 95% CI) | 0.08 [‐1.19, 1.35] |
| 1.9.2 Participant report | 1 | 24 | Mean Difference (IV, Fixed, 95% CI) | 0.08 [‐1.20, 1.36] |
| 1.10 Aggressive behaviour: severity of incidents: Aggression Questionnaire – Hostility subscale (problem‐solving intervention vs control) Show forest plot | 1 | Mean Difference (IV, Fixed, 95% CI) | Subtotals only | |
| 1.10.1 Staff report | 1 | 24 | Mean Difference (IV, Fixed, 95% CI) | 0.75 [‐0.68, 2.18] |
| 1.10.2 Participant report | 1 | 24 | Mean Difference (IV, Fixed, 95% CI) | ‐1.00 [‐84.18, 82.18] |
| 1.11 Aggressive behaviour: severity of incidents: Aggression Questionnaire – Indirect Aggression subscale (problem‐solving vs control) Show forest plot | 1 | Mean Difference (IV, Fixed, 95% CI) | Subtotals only | |
| 1.11.1 Staff report | 1 | 24 | Mean Difference (IV, Fixed, 95% CI) | ‐0.58 [‐2.20, 1.04] |
| 1.11.2 Participant report | 1 | 24 | Mean Difference (IV, Fixed, 95% CI) | ‐2.00 [‐3.33, ‐0.67] |
| 1.12 Aggressive behaviour: severity of incidents: Aggression Questionnaire – Physical Aggression subscale (cognitive re‐appraisal intervention vs control) Show forest plot | 1 | Mean Difference (IV, Fixed, 95% CI) | Subtotals only | |
| 1.12.1 Staff report | 1 | 24 | Mean Difference (IV, Fixed, 95% CI) | ‐1.42 [‐2.92, 0.08] |
| 1.12.2 Participant report | 1 | 24 | Mean Difference (IV, Fixed, 95% CI) | ‐0.08 [‐1.86, 1.70] |
| 1.13 Aggressive behaviour: severity of incidents: Aggression Questionnaire – Verbal Aggression subscale (cognitive re‐appraisal vs control) Show forest plot | 1 | Mean Difference (IV, Fixed, 95% CI) | Subtotals only | |
| 1.13.1 Staff report | 1 | 24 | Mean Difference (IV, Fixed, 95% CI) | 0.75 [‐1.04, 2.54] |
| 1.13.2 Participant report | 1 | 24 | Mean Difference (IV, Fixed, 95% CI) | 0.17 [‐0.81, 1.15] |
| 1.14 Aggressive behaviour: severity of incidents: Aggression Questionnaire – Anger subscale (cognitive re‐appraisal vs control) Show forest plot | 1 | Mean Difference (IV, Fixed, 95% CI) | Subtotals only | |
| 1.14.1 Staff report | 1 | 24 | Mean Difference (IV, Fixed, 95% CI) | 0.50 [‐1.04, 2.04] |
| 1.14.2 Participant report | 1 | 24 | Mean Difference (IV, Fixed, 95% CI) | 0.00 [‐1.21, 1.21] |
| 1.15 Aggressive behaviour: severity of incidents: Aggression Questionnaire – Hostility subscale (cognitive re‐appraisal vs control) Show forest plot | 1 | 48 | Mean Difference (IV, Fixed, 95% CI) | 0.46 [‐0.53, 1.45] |
| 1.15.1 Staff report | 1 | 24 | Mean Difference (IV, Fixed, 95% CI) | 0.58 [‐0.82, 1.98] |
| 1.15.2 Participant report | 1 | 24 | Mean Difference (IV, Fixed, 95% CI) | 0.34 [‐1.06, 1.74] |
| 1.16 Aggressive behaviour: severity of incidents: Aggression Questionnaire – Indirect Aggression subscale (cognitive re‐appraisal vs control) Show forest plot | 1 | Mean Difference (IV, Fixed, 95% CI) | Subtotals only | |
| 1.16.1 Staff report | 1 | 24 | Mean Difference (IV, Fixed, 95% CI) | ‐0.58 [‐2.14, 0.98] |
| 1.16.2 Participant report | 1 | 24 | Mean Difference (IV, Fixed, 95% CI) | ‐0.17 [‐1.46, 1.12] |
| 1.17 Ability to control anger: Provocation Inventory (PI) – self‐report Show forest plot | 3 | Mean Difference (IV, Fixed, 95% CI) | Subtotals only | |
| 1.17.1 Post‐treatment | 3 | 212 | Mean Difference (IV, Fixed, 95% CI) | ‐8.38 [‐14.05, ‐2.71] |
| 1.17.2 At 4‐months follow‐up | 1 | 36 | Mean Difference (IV, Fixed, 95% CI) | ‐4.96 [‐15.82, 5.90] |
| 1.17.3 At 10‐month follow‐up | 1 | 143 | Mean Difference (IV, Fixed, 95% CI) | ‐3.70 [‐10.55, 3.15] |
| 1.18 Ability to control anger: PI – Irritations subscale Show forest plot | 1 | Mean Difference (IV, Fixed, 95% CI) | Subtotals only | |
| 1.18.1 PI – Irritations subscale post‐treatment | 1 | 36 | Mean Difference (IV, Fixed, 95% CI) | ‐2.36 [‐4.67, ‐0.05] |
| 1.18.2 PI – Irritations subscale at 4‐month follow‐up | 1 | 36 | Mean Difference (IV, Fixed, 95% CI) | ‐1.41 [‐3.83, 1.01] |
| 1.19 Ability to control anger: PI – Disrespect subscale Show forest plot | 1 | Mean Difference (IV, Fixed, 95% CI) | Subtotals only | |
| 1.19.1 Post‐treatment | 1 | 36 | Mean Difference (IV, Fixed, 95% CI) | ‐1.16 [‐3.30, 0.98] |
| 1.19.2 At 4‐month follow‐up | 1 | 36 | Mean Difference (IV, Fixed, 95% CI) | ‐0.30 [‐2.82, 2.22] |
| 1.20 Ability to control anger: PI – Unfairness subscale Show forest plot | 1 | Mean Difference (IV, Fixed, 95% CI) | Subtotals only | |
| 1.20.1 Post‐treatment | 1 | 36 | Mean Difference (IV, Fixed, 95% CI) | ‐2.04 [‐4.45, 0.37] |
| 1.20.2 At 4‐month follow‐up | 1 | 36 | Mean Difference (IV, Fixed, 95% CI) | ‐0.19 [‐2.53, 2.15] |
| 1.21 Ability to control anger: PI – Annoying Traits subscale Show forest plot | 1 | Mean Difference (IV, Fixed, 95% CI) | Subtotals only | |
| 1.21.1 Post‐treatment | 1 | 36 | Mean Difference (IV, Fixed, 95% CI) | ‐0.85 [‐4.00, 2.30] |
| 1.21.2 At 4‐month follow‐up | 1 | 36 | Mean Difference (IV, Fixed, 95% CI) | ‐1.55 [‐4.24, 1.14] |
| 1.22 Ability to control anger: PI – Frustrations subscale Show forest plot | 1 | Mean Difference (IV, Fixed, 95% CI) | Subtotals only | |
| 1.22.1 Post‐treatment | 1 | 36 | Mean Difference (IV, Fixed, 95% CI) | ‐0.80 [‐3.95, 2.35] |
| 1.22.2 Ability to control anger: PI – Frustrations subscale at 4‐month follow‐up | 1 | 36 | Mean Difference (IV, Fixed, 95% CI) | ‐1.50 [‐4.20, 1.20] |
| 1.23 Ability to control anger: PI – key worker and carer reports Show forest plot | 2 | Mean Difference (IV, Fixed, 95% CI) | Subtotals only | |
| 1.23.1 Post‐treatment | 2 | 279 | Mean Difference (IV, Fixed, 95% CI) | ‐5.76 [‐9.70, ‐1.82] |
| 1.23.2 At 10‐month follow‐up | 1 | 245 | Mean Difference (IV, Fixed, 95% CI) | ‐4.58 [‐9.27, 0.11] |
| 1.24 Ability to control anger: Profile of Anger Coping Skills (PACS) – self‐report Show forest plot | 1 | Mean Difference (IV, Fixed, 95% CI) | Subtotals only | |
| 1.24.1 Post‐treatment | 1 | 156 | Mean Difference (IV, Fixed, 95% CI) | 8.70 [0.98, 16.42] |
| 1.24.2 At 10‐month follow‐up | 1 | 138 | Mean Difference (IV, Fixed, 95% CI) | 7.70 [‐0.72, 16.12] |
| 1.25 Ability to control anger: PACS – key worker and carer reports Show forest plot | 1 | Mean Difference (IV, Fixed, 95% CI) | Subtotals only | |
| 1.25.1 Post‐treatment | 1 | 260 | Mean Difference (IV, Fixed, 95% CI) | 6.15 [2.08, 10.21] |
| 1.25.2 At 10‐month follow‐up | 1 | 225 | Mean Difference (IV, Fixed, 95% CI) | 5.40 [0.72, 10.09] |
| 1.26 Adaptive functioning: Adaptive Behaviour Scale – Revised, Part II (ABS‐II) Show forest plot | 1 | 28 | Mean Difference (IV, Fixed, 95% CI) | ‐21.73 [‐36.44, ‐7.02] |
| 1.27 Mental state: psychiatric symptoms measured using the Brief Symptom Inventory (BSI) Show forest plot | 1 | 28 | Mean Difference (IV, Fixed, 95% CI) | ‐0.48 [‐0.79, ‐0.17] |
| 1.28 Mental state: psychological distress measured using the Subjective Units of Distress Scale (SUDS) Show forest plot | 1 | 28 | Mean Difference (IV, Fixed, 95% CI) | ‐4.36 [‐6.85, ‐1.87] |
| 1.29 Mental state: depression: Glasgow Depression Scale – Intellectual Disability (GDS‐ID) Show forest plot | 1 | Mean Difference (IV, Fixed, 95% CI) | Subtotals only | |
| 1.29.1 Post‐treatment | 1 | 157 | Mean Difference (IV, Fixed, 95% CI) | ‐0.70 [‐3.06, 1.66] |
| 1.29.2 At 10‐month follow‐up | 1 | 144 | Mean Difference (IV, Fixed, 95% CI) | 0.20 [‐2.15, 2.55] |
| 1.30 Mental state: anxiety: Glasgow Anxiety Scale – Intellectual Disability (GAS‐ID) Show forest plot | 1 | Mean Difference (IV, Fixed, 95% CI) | Subtotals only | |
| 1.30.1 Post‐treatment | 1 | 154 | Mean Difference (IV, Fixed, 95% CI) | ‐2.30 [‐5.39, 0.79] |
| 1.30.2 At 10‐months follow‐up | 1 | 143 | Mean Difference (IV, Fixed, 95% CI) | 0.40 [‐2.59, 3.39] |
| 1.31 Self‐esteem: Rosenberg Self‐Esteem Scale (SES) Show forest plot | 1 | Mean Difference (IV, Fixed, 95% CI) | Subtotals only | |
| 1.31.1 Post‐treatment | 1 | 141 | Mean Difference (IV, Fixed, 95% CI) | 0.30 [‐0.98, 1.58] |
| 1.31.2 At 10‐month follow‐up | 1 | 134 | Mean Difference (IV, Fixed, 95% CI) | ‐0.70 [‐2.22, 0.82] |
| 1.32 Quality of life: Comprehensive Quality of Life Scale: Intellectual Disability (COMQoL‐ID) Show forest plot | 1 | Mean Difference (IV, Fixed, 95% CI) | Subtotals only | |
| 1.32.1 Post‐treatment | 1 | 129 | Mean Difference (IV, Fixed, 95% CI) | ‐5.60 [‐18.11, 6.91] |
| 1.32.2 At 10‐month follow‐up | 1 | 140 | Mean Difference (IV, Fixed, 95% CI) | ‐0.60 [‐13.26, 12.06] |
| 1.33 Costs of service utilisation: Client Service Receipt Inventory (CSRI): cost per person per week of health and social care resource (in British pounds) Show forest plot | 1 | 133 | Mean Difference (IV, Fixed, 95% CI) | 102.99 [‐117.16, 323.14] |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 2.1 Aggressive behaviour: Aberrant Behaviour Checklist (ABC) Total Show forest plot | 2 | Mean Difference (IV, Fixed, 95% CI) | Subtotals only | |
| 2.1.1 At 6 months | 2 | 275 | Mean Difference (IV, Fixed, 95% CI) | ‐7.78 [‐15.23, ‐0.32] |
| 2.1.2 At 12 months | 1 | 225 | Mean Difference (IV, Fixed, 95% CI) | ‐5.20 [‐13.27, 2.87] |
| 2.2 Mental state: Common Mental Disorder (CMD) assessed using the Mini‐Psychiatric Assessment Schedule for Adults with Developmental Disabilities (Mini PAS‐ADD) Show forest plot | 1 | Odds Ratio (M‐H, Fixed, 95% CI) | Subtotals only | |
| 2.2.1 At 6 months | 1 | 214 | Odds Ratio (M‐H, Fixed, 95% CI) | 1.44 [0.83, 2.49] |
| 2.2.2 At 12 months | 1 | 225 | Odds Ratio (M‐H, Fixed, 95% CI) | 0.95 [0.56, 1.62] |
| 2.3 Mental state: severe mental illness (SMI) assessed using the Mini PAS‐ADD Show forest plot | 1 | Odds Ratio (M‐H, Fixed, 95% CI) | Subtotals only | |
| 2.3.1 At 6 months | 1 | 214 | Odds Ratio (M‐H, Fixed, 95% CI) | 1.66 [0.76, 3.62] |
| 2.3.2 At 12 months | 1 | 225 | Odds Ratio (M‐H, Fixed, 95% CI) | 0.87 [0.42, 1.80] |
| 2.4 Mental state: prevalence of affective disorders assessed using the PAS‐ADD checklist Show forest plot | 1 | 60 | Odds Ratio (M‐H, Fixed, 95% CI) | 0.82 [0.24, 2.81] |
| 2.5 Mental state: prevalence of psychotic disorders assessed by the PAS‐ADD checklist Show forest plot | 1 | 60 | Odds Ratio (M‐H, Fixed, 95% CI) | 1.56 [0.24, 10.05] |
| 2.6 Mental state: prevalence of organic mental disorder assessed by the PAS‐ADD Checklist Show forest plot | 1 | 60 | Odds Ratio (M‐H, Fixed, 95% CI) | 0.36 [0.06, 2.01] |
| 2.7 Costs of service utilisation: Client Service receipt Inventory (CSRI) – Total mean health and social care costs per participant Show forest plot | 2 | Mean Difference (IV, Fixed, 95% CI) | Subtotals only | |
| 2.7.1 At 6 months | 1 | 61 | Mean Difference (IV, Fixed, 95% CI) | ‐2200.00 [‐5186.23, 786.23] |
| 2.7.2 At 12 months | 1 | 225 | Mean Difference (IV, Fixed, 95% CI) | ‐448.00 [‐1660.83, 764.83] |
| 2.7.3 At 36 months | 1 | 180 | Mean Difference (IV, Fixed, 95% CI) | 81.00 [‐1610.96, 1772.96] |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 3.1 Aggressive behaviour: frequency of incidents: number of incidents of physical aggression per week during treatment (12 weeks) Show forest plot | 1 | 34 | Mean Difference (IV, Fixed, 95% CI) | ‐2.80 [‐4.37, ‐1.23] |
| 3.2 Aggressive behaviour: frequency of incidents: number of incidents of verbal aggression per week during treatment (12 weeks) Show forest plot | 1 | 34 | Mean Difference (IV, Fixed, 95% CI) | ‐3.30 [‐5.05, ‐1.55] |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 4.1 Aggressive behaviour: aggressive events – carer reports Show forest plot | 2 | 128 | Mean Difference (IV, Random, 95% CI) | ‐10.27 [‐14.86, ‐5.67] |
| 4.2 Aggressive behaviour: staff injuries – carer reports Show forest plot | 1 | 80 | Mean Difference (IV, Fixed, 95% CI) | ‐1.46 [‐1.95, ‐0.97] |
| 4.3 Aggressive behaviour: peer injuries – carer reports Show forest plot | 1 | 80 | Mean Difference (IV, Fixed, 95% CI) | ‐2.47 [‐2.93, ‐2.01] |
| 4.4 Reduction in (additional) medication: emergency medication – carer reports Show forest plot | 2 | 128 | Mean Difference (IV, Random, 95% CI) | ‐3.98 [‐9.68, 1.71] |
| 4.5 Reduction in care needs: physical restraints – carer reports Show forest plot | 2 | 128 | Mean Difference (IV, Random, 95% CI) | ‐7.15 [‐13.16, ‐1.13] |
| 4.6 Reduction in care needs: 1:1 staffing – carer reports Show forest plot | 2 | 128 | Mean Difference (IV, Fixed, 95% CI) | ‐3.94 [‐4.33, ‐3.54] |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 5.1 Ability to control anger: Novaco Anger Scale: Provocation Inventory (NAS:PI) – client ratings Show forest plot | 1 | Mean Difference (IV, Fixed, 95% CI) | Totals not selected | |
| 5.2 Ability to control anger: NAS:PI – carer ratings Show forest plot | 1 | Mean Difference (IV, Fixed, 95% CI) | Totals not selected | |
| 5.3 Mental state: Reiss Screen for Maladaptive Behaviour (REISS) – client ratings Show forest plot | 1 | Mean Difference (IV, Fixed, 95% CI) | Totals not selected | |
| 5.4 Mental state: REISS – carer ratings Show forest plot | 1 | Mean Difference (IV, Fixed, 95% CI) | Totals not selected | |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 6.1 Aggressive behaviour: Aberrant Behaviour Checklist (ABC) – Residential Show forest plot | 1 | Mean Difference (IV, Fixed, 95% CI) | Totals not selected | |
| 6.2 Adaptive functioning: Matson Evaluation of Social Skills for Individuals with Severe Mental Retardation (MESSIER) Show forest plot | 1 | Mean Difference (IV, Fixed, 95% CI) | Totals not selected | |
| 6.3 Adaptive functioning: Unstructured Imitation Assessment (UIA) Show forest plot | 1 | Mean Difference (IV, Fixed, 95% CI) | Totals not selected | |
| 6.4 Adaptive functioning: Motor Imitation Scale (MIS) Show forest plot | 1 | Mean Difference (IV, Fixed, 95% CI) | Totals not selected | |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 7.1 Proportion of participants who achieved a positive response on the ABC (≥ 25% reduction in scores) Show forest plot | 1 | Odds Ratio (M‐H, Fixed, 95% CI) | Totals not selected | |
