TABLE OF CONTENTS

• General information
• Text-word based filtering-how it works
• Text-word based filtering-examples
• Filtering references by MeSH terms

General information

In the reference list we can see all the references assigned to us, with the possibility of studying specific bibliographical data. On the left-hand side, you can filter records according to the following criteria:

• Status: Decision code, i.e. Inclusion/Exclusion criteria 

• Attributes: Year of publication and structured comments

• Text-word terms: Words or phrases present in the title or abstract

• MeSH terms: MeSH terms used to index the record in databases

Text-word based filtering - how it works

Advantages of filters utilisation:

Screening records using clusters created by text-word based filters provides several benefits to the review process:

• training the AI model that drives the prioritization of records,

• efficient identification of the most relevant records,

• quick exclusion of irrelevant records,

• screening becomes more manageable once the similar articles are reviewed within a cluster.

In order to increase the efficiency of your screening process, we recommend the use of filters based on text word terms entered during creating the Instruction, here are two types of text-word terms:

• Positive (green by default) - their presence suggests that the record should be included.

• Negative (red by default) - their presence indicates that the record should be excluded.

TIP: You can select all positive or negative text words within each domain by clicking on 'All positive' or 'All negative'.

TIP: The highlight colour can be customised.

Text-words terms  are organized in a separate domains based on your screening questions. By default, there is a Boolean operator "AND" between each text-word domain, and you cannot change it. It means that when you select "random*" in the study type filters and "Adult" or "Elderly" from the population domain, you will get a set of studies that probably are randomized studies conducted on the adult population. In the single text-word domain, you have the flexibility to choose between "AND" or "OR" operators to connect text-word terms, similar to query formulation in search strategy development.

Detailed information about using operators during the screening are presented in the graphic below. 

Text-word based filtering-examples

You can utilize filters to speed up a Title and abstract screening - Below you will find three examples of how to streamline the screening process:

Exclude obviously irrelevant records with records filtering

Filters can be used to exclude irrelevant records in Focus Mode:

1. Select some negative text word terms (red ones - by default) that indicate studies with e.g. irrelevant populations (paediatric patients - see example) using the Boolean operator "OR".

2. Go to Focus mode with a filtered set of records and make a decision.

3. Send your decisions.

Or to exclude whole batches of definitely irrelevant records in the References view:

1. In the 'search box', add a term that indicates studies with, for example, an irrelevant population, e.g. child*.

2.  Refine your filter to the 'Title' field only, focusing on records containing the selected term in relation to a patient group we are not interested in.

3.  Screen the records in the reference list by title only. 

4. Exclude them using the 'Exclude' (population) button at the bottom of the tool.

5. Send your decisions.

Screen the most relevant records at first

Laser AI filtering features allows also to select records with the highest probability of inclusion:

1.  Select all text-words (desired ones) from each PICOS domain.

2.  Open Focus mode and start screening.

What are the benefits of screening the most relevant records at first?

• Training the AI model to prioritise records

• Fast identifying the most relevant records (which can be screened initially while simultaneously starting full-text screening). 

Tips: To train the model, you must make decisions for at least 32 records and have at least one included, and one excluded record.

Filtering references by MeSH terms

To filter by MeSH terms, go to the "Fields" category at the bottom of the Filters section. The term can be selected in one of two ways: 

• Using the search field to find a term and selecting it from the list of suggestions. Only leaf-terms (the lowest in the hierarchy tree) can be filtered using this method

• By browsing the list of all MeSH terms available for all records in this project. 

In this view you can explore all MeSH terms and perform more extensive filtering, for example by selection terms with a higher hierarchical level than “omega-3” (see “Fatty acids” in the example), you will filter all results that belong to this term, also 

When is it worthwhile to filter references by MeSH terms? 
• To increase sensitivity: Text words may not capture all relevant variations of a concept, or may not be present in the title or abstract at all, potentially missing important records. 
• To increase precision: Text words may have variations or multiple meanings, which may lead to less accurate results. MeSH terms are assigned by human indexers who review each article. MeSH terms help to overcome this problem.

RELATED ARTICLES

• Sorting references  
• Instructions and highlights in the Focus mode for Title and abstract screening  
• Highlight sets  
• How to perform Title and abstract screening  
• Prioritizing records based on their probability of inclusion