AI-Enabled Screening Workflows
Learn how to make the most of AI tags in MLM-AI
Last updated
Learn how to make the most of AI tags in MLM-AI
Last updated
Biologit MLM-AI supports various strategies to improve quality and speed up screening workflows using AI features. this can be achieved by focusing efforts on the results more likely to contain relevant safety information.
In this document we explore common AI-based screening strategies with examples.
Every abstract retrieved by Biologit MLM-AI can receive AI tags representing predictions that correspond to relevant safety surveillance information:
Potential safety event,
Special situations,
Identifiable patients,
Animal or In-vitro study
The diagram below illustrates how AI tags helps users focus on the most relevant results. Consider for example an ICSR pharmacovigilance workflow focusing on adverse events on humans. From all results retrieved, users can prioritize the ones with an identifiable patient (patient tag), progressively moving to the abstracts tagged with a suspect adverse event (suspect AE).
The following AI Tags are available in MLM-AI:
Suspected AE
The abstract or full text may contain the description of an adverse event. Thus, MLM-AI flagged the abstract as a candidate for further screening.
Suspected Case
The abstract or full text may contain the description of a case (identifiable patient).
Animal/In-Vitro
The abstract describes animal or in-vitro studies.
Special Situations
Any detected special situation (elderly, pediatric, pregnancy) is also flagged as a tag.
Identifiable Patient
A confirmed patient mention (ex: "a 72-year old woman" etc)
With AI tags, users can:
Prioritize abstracts classified as Suspected AE, by ranking or filtering results.
Use less resource intensive screening methods where safety events are unlikely. For example:
This section contains examples of how to employ AI tags for faster screening.
The ranking is preserved once users start reviewing articles in the detail page: users will screen the prioritized articles first.
Abstracts not tagged can in principle receive less scrutiny during reviews. Examples of how this can be done:
Update your screening process so that the "Not Suspected AE" abstracts receive fewer QC steps (ie. "sampled" QC)
For example: In an ICSR workflow, Batch Review can be used to refute all animal/in vitro studies at once (typically not a valid ICSR):
From the search box, select articles tagged as Animal/In Vitro with tag:animal
Then click "Batch Review" to quickly review all abstracts from the same screen, according to the search criteria specified
Use the checkbox to select irrelevant articles and save your decision for all selections at once
Users can further reduce effort by configuring monitors to automatically pre-screen abstracts based on AI tags.
All pre-screened articles remain visible in the results. They are presented in the "Reviewed" tab and can still be reviewed as part of the quality control process.
of Non Suspected AE abstracts.
Finally, users can articles by Tag, skipping screening and perform only QC.
Ensure automated screening using AI Tags is implemented with adequate risk-based quality controls in place for your scenario. and can help with your requirements.
Abstracts in a review can be ranked according to their AI Tags. In the simply sort the abstracts by a tag such as "Suspected AE" or "Identifiable Patient":
These can be retrieved from any review by searching notag:adverse(see more details on )
Screen quicker with
enables faster screening of many abstracts at once.
See also and page.
In the users can select the level of pre-screening for a given Monitor according to which AI tags are present in an article:
See also for more details on this feature.
