Canada Spends Logo

Explore our Budget 2026 analysis

Grant: Master's Award: Canada Graduate Research Scholarships - Master's program

McMaster University

$27,000.00

Evaluating Artificial Intelligence Assisted Screening in Clinical Practice Guideline Development

FY 2025-26

Summary
Clinical practice guidelines rely on rigorous evidence reviews, but screening large volumes of literature is time-consuming and resource intensive. AI-assisted screening tools are promoted to reduce workload by prioritizing citations, yet it is uncertain whether they may mis studies that influence guideline conclusions. No prospective research has examined how AI-assisted screening affects the final evidence base or the recommendations made by guidance panels, leaving a methodological gap. This project will assess whether AI-assisted title and abstract screening leads to differences in included studies and guideline recommendations compared with full human screening. DistillerSR and WPPI-Reviewer will be evaluated at sensitivity thresholds of 95% and 200%. Human-screened citation sets from guideline projects led by the Program of Evidence-Based Care (PEBC) will serve as the reference standard. AI tools will be trained on a subset of screened citations, and simulated AI-assisted workflows will be generated. Evidence sets produced at each sensitivity level will be compared with those identified through full human screening to determine whether key studies are missed. If fewer studies are found at 95% sensitivity, guideline panel members will create recommendations using both the AI-derived and human derived evidence sets to assess whether any recommendations differ. The study will also compare tool performance, workload savings, and patterns in missed studies. This research will provide a prospective evaluation of whether AI-assisted screening can be safely integrated into guideline development in practice. Findings will clarify when AI screening is appropriate, when full human review is necessary, and how these tools should be used responsibly in evidence synthesis.
Program
Master's Award: Canada Graduate Research Scholarships - Master's program
Location
Ontario, Canada
Recipient Type
Academia
Type
Grant
Source
CIHR Funding Decisions Database

As published

Id
572592
Name
Low Ashley
Region
Ontario
Country
Canada
Orgtype
Institution Paid; Research Institution
Prcname2
Special Cases - Awards Programs
Projecttitle
Evaluating Artificial Intelligence Assisted Screening in Clinical Practice Guideline Development
Programtype2
Studentship Programs
Orgnameinp2
McMaster University
Cihramount2
$27,000
Cihrequipment2
$0
Programname2
Master's Award: Canada Graduate Research Scholarships - Master's program
Keyworddelim
Accuracy; Artificial Intelligence; Clinical Practice Guidelines; Distillersr; Eppi-Reviewer; Systematic Review; Workload Savings
Abstract
Clinical practice guidelines rely on rigorous evidence reviews, but screening large volumes of literature is time-consuming and resource intensive. AI-assisted screening tools are promoted to reduce workload by prioritizing citations, yet it is uncertain whether they may mis studies that influence guideline conclusions. No prospective research has examined how AI-assisted screening affects the final evidence base or the recommendations made by guidance panels, leaving a methodological gap. This project will assess whether AI-assisted title and abstract screening leads to differences in included studies and guideline recommendations compared with full human screening. DistillerSR and WPPI-Reviewer will be evaluated at sensitivity thresholds of 95% and 200%. Human-screened citation sets from guideline projects led by the Program of Evidence-Based Care (PEBC) will serve as the reference standard. AI tools will be trained on a subset of screened citations, and simulated AI-assisted workflows will be generated. Evidence sets produced at each sensitivity level will be compared with those identified through full human screening to determine whether key studies are missed. If fewer studies are found at 95% sensitivity, guideline panel members will create recommendations using both the AI-derived and human derived evidence sets to assess whether any recommendations differ. The study will also compare tool performance, workload savings, and patterns in missed studies. This research will provide a prospective evaluation of whether AI-assisted screening can be safely integrated into guideline development in practice. Findings will clarify when AI screening is appropriate, when full human review is necessary, and how these tools should be used responsibly in evidence synthesis.
Competitiondate
202512
Approvedterm2
1 yr 0 mth
Pinamesdelim
Low, Ashley
Approvedterm
1 an 0 mois
Cihrcontribution2
~ $27,000~ $0~