TABLE OF CONTENTS

• Scope of Use for This Data Extraction Form
• Overview of the Data Extraction Form
• Data extraction process
  • Database-format extraction
  • Connections between PDF and values in data extraction forms
  • Models suggestions
  • Vocabulary fields
• Tabs description
  • Study details tab
  • Intervention tab
  • Population tab
  • Results tab
    • Standard 2×2 contingency table
    • Accuracy measure
    • Information on the broader implementation and acceptability
    • Impact of tests on patient outcome

Scope of Use for This Data Extraction Form

This data extraction form is designed for use in reviews that aim to:

Assess the effectiveness of diagnostic technologie (e.g., in vitro diagnostic tests, imaging tests, etc.), including:

• Diagnostic accuracy
• Implementation and acceptability issues related to the test, such as inter-rater reliability, adverse events, acceptability, etc.
• Impact on patient outcomes, including morbidity, mortality, and health-related quality of life (HRQoL)

Overview of the Data Extraction Form

• Three tabs are dedicated to extracting data related to the study methodology, including:

  • Study details

  • Details about the intervention

  • Information on the study population

• One tab is dedicated to extracting outcome data.

In our template libraries, you will also find data extraction forms for risk of bias (ROB) assessment, including QUADAS-2, which is designed for both single-test accuracy studies and comparative accuracy studies. You can import this form directly into your project for use

Data extraction process

Database-format extraction

Please note that Laser AI supports a database-driven approach to data extraction. This means, for example, that different diagnostic accuracy data are extracted in separate sections. You won’t find dedicated sections for sensitivity orspecificty—instead, there is a single section called Accuracy measures. Within this section, you should select the specific output you are extracting (e.g., sensitivity in the first subsection, specificity in the second), using the controlled vocabulary.

Read more about this database approach 

Connections between PDF and values in data extraction forms

The best way to extract data in Laser AI is by using the highlighting mechanism. This method makes the quality assurance process easier, as clicking on a field with an extracted value redirects you to the corresponding section in the PDF where the value was extracted. To use this mechanism, you need to:

1.  Locate the specific data in the document and highlight it.
2.  Click the 'Plus' button (when empty data extraction fields exist)  to add the highlighted data to the corresponding data extraction field. In this way the relation between data in PDF and value in data extraction form will be established. When there are no empty fields in data extraction form, click the 'Plus' button located below the data extraction fields. The tool will create additional field and the extracted value will be automatically added to the data extraction form. Please note that this option is available only for sections with a single data extraction field 

Models suggestions

You may notice that some fields are coloured yellow, which indicates that an AI model has been enabled for that field. Your task is to check the suggestion and either accept or reject it. After selecting the relevant field, you will be taken to the part of the text where the suggestion has been made so that you can check its accuracy. Next, accept or reject the suggestion in the data extraction form.

Vocabulary fields

Most of the data extraction fields are text fields, such as those used for extracting numerical values. However, some sections include vocabulary fields, where reviewers select data from a predefined list. During extraction, each reviewer has permission to create new codes if needed.

Tabs description

Study details tab

This tab is dedicated to extracting study characteristics, such as study type, funding source, country, and more. 

Sometime below section you will find fields called subsections, where you can extract concepts than appear more than once in study for example country. This section can be repeated if the study includes more than one country To do this, click the 'Add' button. More details about subsections are available here.  

Intervention tab

This tab is dedicated to extracting data related to interventions, including both the index test and the reference test. Data for the reference test (gold standard) are extracted in separate sections, and the list of extraction fields differs slightly from those used for the index test.

Please note that the value entered in the “Index Test Name” field (in the index test section) will be used as a label during the extraction of outcome data.

Population tab

In this tab, you can extract data for the study population, including:

• Patient sampling methods
  
• Target population
  
• Details about the study population, such as eligibility criteria
  
• Patient flow (number of patients throughout the study)
  
• Baseline characteristics

In the Patient Flow and Patient Characteristics sections, you will find subsections for each study group, allowing you to extract data for each index test. The fields for extraction are generated automatically once you enter the Index Test Name in the intervention tab.

The recommended method of extraction is semi-automated table extraction, where possible. Read more about it in  this article.  

Results tab

This tab includes four sections dedicated to extracting different types of outcomes:

• Accuracy measures

• Standard 2×2 contingency table 

• Outcomes related to index test implementation and acceptability

• Data on the impact of the index test on patients (e.g., morbidity, quality of life, etc.)

Standard 2×2 contingency table

In this section, you can extract values for true positives (TP), false positives (FP), false negatives (FN), and true negatives (TN).

First, select the unit of measurement from the list (e.g., person-based, lesion-based, or sample-based).Next, select the type of measure from the list.

Data should be extracted separately for each index test.

The recommended method of extraction is semi-automated table extraction, where possible. Read more about it in  this article.  

Accuracy measure

This section is organized in a similar way to the previous one. You should extract each accuracy measure as a separate entry. Relevant vocabularies for different types of accuracy measures are provided to guide the extraction process.

Information on the broader implementation and acceptability

In this section, you can extract additional outcomes related to index test implementation or patient acceptability.

In the first field, select the outcome type from the provided list.

Please note that you can create a new code for a new outcome if it is not listed.

Impact of tests on patient outcome

In this section, you can extract additional clinical outcomes such as Quality of life data. 

In the first field, select the outcome type from the provided list.

Please note that you can create a new code for a new outcome if it is not listed