Article type
Year
Abstract
Background:
Identifying all evidence relevant to a systematic review remains a critical yet time-consuming step in the evidence synthesis process. Machine learning methods and collaborative screening software constitute potential means of reducing the workload necessary to perform citation screening, without sacrificing rigor. But while sophisticated web-based tools for performing citation screening have emerged (e.g., Covidence, Rayyan) they tend to feature relatively limited automation functionality. Conversely, open-source research prototypes (e.g., abstrackr, RobotReviewer) offer more sophisticated machine learning to aid synthesis but lack simple user experiences.
Objectives:
We introduce PICOPortal, a new web-based tool for citation screening that facilitates systematic reviews that aspires to combine the strengths of a modern user interface and cutting-edge machine learning functionality. Users can create publicly available profiles within PICOPortal that include their areas of expertise and levels of experience, potentially facilitating collaboration across systematic review groups based in other labs or research organizations around the world. The tool is free for academic users.
Methods:
PICOPortal provides project management and basic reference management functionality in a modern user interface along with state-of-the-art machine learning capabilities to facilitate an efficient review. It integrates a validated, state-of-the-art machine learning to optionally including only randomized controlled trial (RCT), it detects de-duplication of citations and automatically extracts snippets of text from titles & abstracts pertaining to the descriptions of trial Populations, Interventions/Comparators, and Outcomes (PICO elements), respectively (Figure 1).
Additionally, extraction of these snippets facilitates automated topic-scope based exclusion of articles on the explicit basis of one or more elements such as Population that is inappropriate for the scope of the review at hand. This is in contrast to models that make an overall relevance determination without explicit reference to an underlying PICO criterion.
Conclusions:
PICOPortal is a new web-based tool for collaborative citation screening for systematic reviews. It features cutting-edge machine learning models that are integrated into an intuitive interface, thus combining the respective strengths of existing commercial and academic citation screening tools. PICOPortal is designed to support a team’s systematic review process through it’s entire life cycle; while maintaining an emphasis on academic rigor, workflow optimization and flexibility, and global collaboration. PICOPortal is free for academic users.
Patient or healthcare consumer involvement:
Systematic reviews provide the best means of realizing the practice of evidence-based medicine (EBM). Citation screening, which the described tool facilitates, is a key component of such reviews. Patients, therefore, stand to benefit indirectly from researcher use of the PICOPortal tool described in this abstract.
Identifying all evidence relevant to a systematic review remains a critical yet time-consuming step in the evidence synthesis process. Machine learning methods and collaborative screening software constitute potential means of reducing the workload necessary to perform citation screening, without sacrificing rigor. But while sophisticated web-based tools for performing citation screening have emerged (e.g., Covidence, Rayyan) they tend to feature relatively limited automation functionality. Conversely, open-source research prototypes (e.g., abstrackr, RobotReviewer) offer more sophisticated machine learning to aid synthesis but lack simple user experiences.
Objectives:
We introduce PICOPortal, a new web-based tool for citation screening that facilitates systematic reviews that aspires to combine the strengths of a modern user interface and cutting-edge machine learning functionality. Users can create publicly available profiles within PICOPortal that include their areas of expertise and levels of experience, potentially facilitating collaboration across systematic review groups based in other labs or research organizations around the world. The tool is free for academic users.
Methods:
PICOPortal provides project management and basic reference management functionality in a modern user interface along with state-of-the-art machine learning capabilities to facilitate an efficient review. It integrates a validated, state-of-the-art machine learning to optionally including only randomized controlled trial (RCT), it detects de-duplication of citations and automatically extracts snippets of text from titles & abstracts pertaining to the descriptions of trial Populations, Interventions/Comparators, and Outcomes (PICO elements), respectively (Figure 1).
Additionally, extraction of these snippets facilitates automated topic-scope based exclusion of articles on the explicit basis of one or more elements such as Population that is inappropriate for the scope of the review at hand. This is in contrast to models that make an overall relevance determination without explicit reference to an underlying PICO criterion.
Conclusions:
PICOPortal is a new web-based tool for collaborative citation screening for systematic reviews. It features cutting-edge machine learning models that are integrated into an intuitive interface, thus combining the respective strengths of existing commercial and academic citation screening tools. PICOPortal is designed to support a team’s systematic review process through it’s entire life cycle; while maintaining an emphasis on academic rigor, workflow optimization and flexibility, and global collaboration. PICOPortal is free for academic users.
Patient or healthcare consumer involvement:
Systematic reviews provide the best means of realizing the practice of evidence-based medicine (EBM). Citation screening, which the described tool facilitates, is a key component of such reviews. Patients, therefore, stand to benefit indirectly from researcher use of the PICOPortal tool described in this abstract.