LabARIA: A Multi-Agent System for Integrative Analytics in Clinical Laboratories

Background Laboratory medicine underpins clinical decision-making across nearly every care setting. Clinical laboratory errors and delays can compromise diagnostic reliability and patient care. However, laboratory quality and operational data are commonly reviewed through separate manual workflows, limiting timely recognition of emerging problems across complementary data sources. Methods We developed and evaluated LabARIA (Laboratory Agentic Reasoning and Integrative Analytics), a multi-agent AI system for coordinated surveillance of specimen workflows, analytical quality control, and laboratory operations. We conducted a single-center retrospective system development and evaluation at NewYork-Presbyterian Hospital/Weill Cornell Medical Center using laboratory data collected from May 1 through June 18, 2026, including 177,480 specimen-panel records, 75,103 internal quality-control (IQC) measurements, 633 proficiency-testing challenges, and 422 recalibration or maintenance actions. Forecasting agents were evaluated using patient-result streams with controlled perturbations, and synthetic staffing and inventory data. We assessed specimen rejection, cancellation, and turnaround-time patterns; warning recall and lead time before documented laboratory actions; assay-specific patient-based real-time quality control (PBRTQC) performance; and exploratory operational forecasting discrimination. Findings LabARIA characterized specimen and workflow patterns across clinical units and time periods, including hemolysis-related potassium rejections in the intensive care unit during early morning hours (49 of 367 tests [13.3%]). For analytical quality monitoring, the IQC agent generated warnings before 407 of 422 recorded QC actions (recall, 0.964 [95% CI, 0.942-0.978]), with a median lead time of 5.40 days (95% CI, 3.75-6.07). PBRTQC agent selected different monitoring strategies across 35 analytes. In a separate analysis, adding PBRTQC increased action recall from 0.964 to 0.993 (95% CI, 0.979-0.998). A dashboard presented tiered alerts for human review, and an accompanying chatbot supported interactive exploration of the summaries. For the operational forecasting module, areas under the receiver operating characteristic curve for 7-day inventory and staffing risk forecasting were 0.800 (95% CI, 0.696-0.881) and 0.877 (95% CI, 0.778-0.950), respectively. Interpretation LabARIA demonstrated the feasibility of coordinating laboratory quality surveillance across heterogeneous data sources, providing a foundation for intelligent clinical laboratories. Prospective studies are warranted to determine whether this approach improves monitoring beyond conventional methods, supports laboratory decisions, and benefits patient care without excessive alert burden.

Authors

Institutions

Publication Details

Journal
medRxiv
Published
2026-11-15
DOI
https://doi.org/10.64898/2026.11.12.26367970
Primary Topic
Clinical Laboratory Practices and Quality Control
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

LabARIA: A Multi-Agent System for Integrative Analytics in Clinical Laboratories

Melissa M. Cushing, Weishen Pan, Zhenxing Xu, He S. Yang et al.
medRxiv
Clinical Laboratory Practices and Quality Control
preprint

LabARIA: A Multi-Agent System for Integrative Analytics in Clinical Laboratories

Melissa M. Cushing, Weishen Pan, Zhenxing Xu, He S. Yang, Haoyang Li, Zhen Zhao, Yaxin Li, Fei Wang
preprint en

Abstract

Background Laboratory medicine underpins clinical decision-making across nearly every care setting. Clinical laboratory errors and delays can compromise diagnostic reliability and patient care. However, laboratory quality and operational data are commonly reviewed through separate manual workflows, limiting timely recognition of emerging problems across complementary data sources. Methods We developed and evaluated LabARIA (Laboratory Agentic Reasoning and Integrative Analytics), a multi-agent AI system for coordinated surveillance of specimen workflows, analytical quality control, and laboratory operations. We conducted a single-center retrospective system development and evaluation at NewYork-Presbyterian Hospital/Weill Cornell Medical Center using laboratory data collected from May 1 through June 18, 2026, including 177,480 specimen-panel records, 75,103 internal quality-control (IQC) measurements, 633 proficiency-testing challenges, and 422 recalibration or maintenance actions. Forecasting agents were evaluated using patient-result streams with controlled perturbations, and synthetic staffing and inventory data. We assessed specimen rejection, cancellation, and turnaround-time patterns; warning recall and lead time before documented laboratory actions; assay-specific patient-based real-time quality control (PBRTQC) performance; and exploratory operational forecasting discrimination. Findings LabARIA characterized specimen and workflow patterns across clinical units and time periods, including hemolysis-related potassium rejections in the intensive care unit during early morning hours (49 of 367 tests [13.3%]). For analytical quality monitoring, the IQC agent generated warnings before 407 of 422 recorded QC actions (recall, 0.964 [95% CI, 0.942-0.978]), with a median lead time of 5.40 days (95% CI, 3.75-6.07). PBRTQC agent selected different monitoring strategies across 35 analytes. In a separate analysis, adding PBRTQC increased action recall from 0.964 to 0.993 (95% CI, 0.979-0.998). A dashboard presented tiered alerts for human review, and an accompanying chatbot supported interactive exploration of the summaries. For the operational forecasting module, areas under the receiver operating characteristic curve for 7-day inventory and staffing risk forecasting were 0.800 (95% CI, 0.696-0.881) and 0.877 (95% CI, 0.778-0.950), respectively. Interpretation LabARIA demonstrated the feasibility of coordinating laboratory quality surveillance across heterogeneous data sources, providing a foundation for intelligent clinical laboratories. Prospective studies are warranted to determine whether this approach improves monitoring beyond conventional methods, supports laboratory decisions, and benefits patient care without excessive alert burden.

medRxiv
Weill Cornell Medicine (US)
Clinical Laboratory Practices and Quality Control
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon