Multi-Agent Collaborative ADR Prediction With Human-Machine Decision Comparison
Trial statusRecruiting
Trial phaseNot listed
Trial typeObservational
Biological sexAll
AgeNot listed
SponsorPeking University Third Hospital
About this trial
This study develops a multi-agent collaborative prediction model to forecast adverse drug reactions using real-world clinical medical records. It validates model performance via evidence-based data and compares decision outputs between the AI model and clinical physicians, aiming to improve early identification of drug adverse events. Only de-identified historical medical data will be analyzed; no new clinical interventions will be conducted, with no additional risks to participants.
Eligibility criteria
This trial does not accept healthy volunteersQualifiers
Cases shall involve drug categories including anti-infectives, cardiovascular agents, anti-tumor drugs, central nervous system drugs, digestive system drugs, etc. Each case must contain at least one definite adverse drug reaction (ADR) event, with complete supporting documentation (medical history, medication history, ADR occurrence process, and clinical outcome).
Disqualifiers
Cases with incomplete supporting documentation lacking medical history, medication history, ADR occurrence process or clinical outcome.
Cases only with suspected or possible ADRs without definite clinical confirmation.
Trial population
This study includes two types of research subjects: Retrospective de-identified adverse drug reaction (ADR) medical records: A total of 253 ADR consultation cases covering anti-infectives, cardiovascular agents, anti-tumor drugs, central nervous system drugs and digestive system drugs. Each case contains complete medical history, medication records, ADR occurrence process and clinical outcome data, as well as at least one clinically confirmed definite ADR event. Clinical evaluators: 20 licensed physicians or pharmacists holding intermediate or higher professional titles, with clinical pharmacy practice and regular participation in hospital ADR monitoring and consultation work.
Trial design
Design model
Other
Time perspective
Other
Treatments tested in this trial
Retrospective medical record data analysis only, no clinical intervention
Other interventionThis study only analyzes de-identified historical electronic medical record data to build a multi-agent AI prediction model for adverse drug reactions. No drugs, medical devices, or clinical treatment interventions will be applied to any human subjects.
Treatment groups
No treatment groups listed
Trial outcomes
Primary outcomes
1
Coverage rate of known ADRs
Time frame
Up to 8 weeks2
Objective question accuracy
Time frame
Up to 24 weeks3
Concordance rate of predicted unknown ADRs
Time frame
Up to 8 weeks4
Expert-rated subjective answer quality
Time frame
Up to 24 weeksSecondary outcomes
1
Subgroup differences in ADR recognition coverage rate
Time frame
Up to 24 weeks2
Inter-rater consistency
Time frame
Up to 24 weeks3
Subgroup differences in answer quality score
Time frame
Up to 24 weeks4
Rater acceptance scale score
Time frame
Up to 24 weeksSponsors and contacts
Click on the lead sponsor to view all of their trials.
Source ClinicalTrials.gov