Deep Attention Model for Triage of Emergency Department Patients
summary
A deep model predicts the number of ED resources a patient will need by combining structured fields and free‑text nurse notes, using word attention over embedded tokens and a BLSTM encoder; trained on ~338,500 visits, it reaches AUC ~88% for identifying resource‑intensive patients and ~44% accuracy for exact resource category, yields about a 16 percentage‑point lift versus nurses, highlights keywords for interpretability, and runs inference in seconds for deployment.
AI-written summary · about
chapters (17)
- 0:00Intro
- 0:35Presentation
- 1:15Outline
- 1:42Search Process
- 6:32Data
- 8:20Text Data
- 8:39LowDimensional Embedding
- 10:01Sequences
- 11:56Deep Learning
- 13:51Summarization
- 16:00Deep Wide Models
- 17:02Methodology
- 17:43Research Question 1
- 18:41Research Question 2
- 22:00Implementation
- 24:13Summary
- 24:51Questions
related talks
20mBERT Model for Real World Healthcare Data
32mAnalyzing social media for suicide risk using natural language processing
05mHow generative AI can support medical staff
34mExtreme Multilabel Classification in the Biomedical NLP Domain
20mAugmenting Mental Health Care in the Digital Age: Machine Learning as a Therapist Assistant
31mBig data for tiny patients: Applying ML to pediatrics
28mAI in healthcare - Deep learning for precision psychiatry
16mMultimodal AI: Transforming Patient-Centered Healthcare