Canonical text (the exact bytes fingerprinted)

TRUVACE RECORD VERSION
record: TRV-2026-0618
version: 1
kind: certified
reason: Certified into the record
timestamp: 2026-08-01T06:08:30.259375Z
status: published
lens: g_space
sector: health
headline: The Sydney Triage to Admission Risk Tool With Artificial Intelligence (START-AI) to Support Decision Making in Emergency Departments: Model Explainability and Feature Importance Analysis
dek: Objective Evaluate the importance of specific variables contributing to a recently reported Artificial Intelligence (AI) prediction model called Sydney Triage to Admission Risk Tool with Artificial Intelligence (START-AI) to predict inpatient admission from the Emergency Department (ED). Methods A model explainability analysis was undertaken using single-centre ED electronic medical record data over 2 years. The START-AI model, which comprises ensemble machine learning and a transformer-based algorithm to enhanc…
gain_title: START-AI improved prediction of inpatient admission from the emergency department from AUROC 0.78 to 0.90 when ensemble and transformer features were added.
problem_title: (none)
trace_subject: (none)
gain_reading: START-AI improved prediction of inpatient admission from the emergency department from AUROC 0.78 to 0.90 when ensemble and transformer features were added.
gain_evidence: reaching an AUROC of 0.90 (95% CI 0.89, 0.90) after all features of the START-AI tool were added sequentially | original START tool alone had an AUROC of 0.78 (95% CI 0.77, 0.78) for prediction of inpatient admission
problem_reading: (none)
problem_evidence: (none)
quick_read: In a single-centre study using two years of ED electronic records, investigators evaluated the Sydney Triage to Admission Risk Tool with Artificial Intelligence (START-AI) by adding features one by one and calculating permutation importance and SHAP values for inpatient admission prediction.

Understanding which inputs most improve AUROC matters for safe deployment, trust, and workflow integration in emergency care, but the analysis remains retrospective and single-centre, leaving prospective validation and impact on actual admission decisions uncertain.
limitation: Findings are based on single-centre data, limiting generalizability to other ED settings.
tag: Evidence-backed gain
key_points: Analysis used single-centre ED electronic medical record data over 2 years to re-run START-AI with sequential feature addition. | Largest stepwise AUROC gains came from triage comments, ED case history notes, any blood test result especially lactate and C-reactive protein, and any CT order. | Overall feature importance was highest for presence of any blood test result, START score, and C-reactive protein, while vital signs did not show stepwise AUROC increase.
rundown: Researchers re-ran the START-AI model, which combines ensemble machine learning and a transformer-based algorithm, adding each feature sequentially and measuring change in AUROC and SHAP values.

The explainability results identified triage comments, case history notes, blood tests and CT orders as drivers of performance, informing how the tool can be further developed and deployed in clinical settings.
sources:
- peer_reviewed | Emergency Medicine Australasia | https://doi.org/10.1111/1742-6723.70285 | 2026-08-01
prev: 0000000000000000000000000000000000000000000000000000000000000000

sha256

372fc37694faa4c23becaaa1ef4dfcbe30255421bd87c183d126eb27fe9d9433

previous

0000000000000000000000000000000000000000000000000000000000000000