What Works?
  • What is What Works?

    >

  • Who is What Works for?

    >

  • Why should I use What Works?

    >

  • How do I use What Works?

    >

  • How do I submit an intervention?

    >

  • Get involved in What Works

    >

  • Understanding the evidence base

    >

  • The SEND Code of Practice

    >

  • Glossary of terms

    >

  • What Works Training Database

    >


Description of the evidence levels 

An intervention's level of evidence is the extent to which there is data available to support a specific intervention. Such data needs to be publicly available in the published literature or on websites. There are three levels:

  • Strong - this includes at least one positive systematic review plus subsequent trials as available
  • Moderate - this would include single randomised controlled studies or quasi-experimental studies
  • Indicative - this means good face validity but limited research evidence i.e. case studies or 'before and after' studies

Below you can find the specific criteria for each of the three levels of evidence. For more explanation of the criteria, see our guide

 

Strong

Evidence will be graded as strong if it includes at least one systematic review plus subsequent trials as available. Systematic reviews are a high level overview of research around a specific research question, that aim to gather the evidence related to that research question in order to answer it. 'Strong' evidence will demonstrate a review of previous literature and evidence and will also show evidence from further trials.

For example, a recent systematic review of stuttering treatments found that 6 out of 13 treatments for children who stammer used studies from evidence behind the Lidcombe Programme.

 

Moderate

Evidence will be graded as moderate if it meets one of the following criteria:

1. Single (or multiple) randomised controlled trial- this level of evaluation would include a random assignment of children or young people to groups; a control group and an intervention group.

For example, the Talk Boost intervention was evaluated using 160 children who were allocated randomly to an intervention or a control group.

2. Quasi-experimental study-this type of experiment compares different groups of participants on some measure, with group membership usually manipulated by the researcher. Participants aren't randomly allocated to groups. For example, the evidence associated with the Naturalistic Speech Intelligibility intervention cites quasi experimental design in researching this intervention for children with speech and language impairments.

 

Indicative

Evidence will be graded as indicative if results have been analysed using appropriate statistics to show a statistically significant difference using one of the following research measures:

1. Control and targeted items are measured in a before and after design- this would be a research design showing baseline measures prior to an intervention or approach and outcome measures after an intervention or approach.

2. Recently standardised assessments are used as comparison and control measure (standardised on an appropriate sample) in a before and after design.Standardised assessments are accepted by the moderating group as away to show changes within a before and after research study.

 

And one of the following research designs:

3. A case series with a multiple baseline design- a case series tracks the progress of a group of children with a similar need during an intervention e.g. phonology therapy. A multiple baseline design means that the intervention would be staggered, for example all children would have an initial baseline phonology assessment completed, then they would all begin their therapy at different times. If participants showed a significant change in their speech across the before and after measures, with the changes seen during the period of the intervention, then this infers that the treatment works and change is not attributed to another factor.

4. A matched control group is used in a group comparison (sample size large enough to calculate an effect size)- where a control group is used, the groupneeds to be matched. For example, a control group may be children of a similarage, with similar speech, language and communication needs.

 

This paper, which contributes to the evidence behind Palin Parent Child Interaction Therapy shows examples of all of the above indicative criteria.

 

 

N.b. Descriptive case studies without any experimental control will not be accepted onto the database, but may be submitted to be included in our 'in practice' section, which consists of case studies associated with included interventions. You'll notice that all the above evaluation methods have some element of control. This allows us to have more certainty that it is an intervention or an approach that is having an effect, and not any other factor within a child or young person's life.