Syst Rev June 2023 PMID 37340494

Performance of active learning models for screening prioritization in systematic reviews: a simulation study into the Average Time to Discover relevant records.

Gerbrich Ferdinands, Raoul Schram, Jonathan de Bruin, Ayoub Bagheri, Daniel L Oberski, Lars Tummers, Jelle Jasper Teijema, Rens van de Schoot

Full text PDF Download social card

Mentions

Peak: 1 in Jun 2023

across 3 sources since Jun 2023

Extended reach

4,116

potential audience

Total likes

Video views

Mentions per month

Jun 2023 – Oct 2023

Where it's discussed

Blogs2 articles
X / Twitter1 posts
YouTube1 videos

AI synopsis

What people are saying

The discussions showcase a keen interest in how active learning models can reduce screening time while maintaining accuracy, with phrases like 'game changer' and 'promising approach' emphasizing their perceived transformative potential. Posts often reference specific technical aspects, such as metrics (ATDBR) and models, reflecting active engagement and recognition of the study's practical impact.

AgreementModerate agreement

Most discussions acknowledge the positive findings of the publication, highlighting the effectiveness of active learning models in speeding up systematic reviews.

InterestHigh interest

Posts demonstrate high interest by exploring implications and potential applications, with some emphasizing how this research could transform review processes.

EngagementModerate engagement

Comments include references to specific models like support vector machines and metrics such as Time to Discovery, indicating thoughtful engagement.

ImpactModerate impact

Several discussions suggest that this study could significantly influence future practices in systematic reviews and research methodologies.

Abstract in brief

This study compares different active learning models and techniques (like naive Bayes, logistic regression, support vector machines, and random forest combined with TFIDF or doc2vec) to see how effectively they can speed up the process of screening records in systematic reviews, while still finding most relevant publications.
The models can cut down the number of records that need to be screened by about 91-92% but still identify 95% of all relevant records, using new metrics like Time to Discovery (TD) and Average Time to Discovery (ATDBR) to measure how quickly relevant records are found.
Overall, active learning tools show strong promise in making systematic reviews faster and less labor-intensive by prioritizing records likely to be relevant early in the screening process.]

Read the full abstract

All mentions

Active Learning Models Enhance Systematic Review Screening Efficiency

This study compares active learning models like naive Bayes, logistic regression, support vector machines, and random forest to improve screening in systematic reviews. The models significantly reduce screening effort…

Medicine RTCL TV · Oct 6, 2023 29 views

General Medicine and Medical Evidence @ BMC

@MedicalEvidence · X

Performance of active learning models for screening prioritization in systematic reviews: a simulation study into the Average Time to Discover relevant records https://t.co/blhRKNZ7Cd

Jun 20, 2023 · 4,088 followers View post →