Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

TriageSim: A Conversational Emergency Triage Simulation Framework from Structured Electronic Health Records

Dipankar Srirag, Quoc Dung Nguyen, Aditya Joshi, Padmanesan Narasimhan, Salil Kanhere

TL;DR

TriageSim is introduced, a simulation framework for generating persona-conditioned triage conversations from structured records that enables multi-turn nurse-patient interactions with explicit control over disfluency and decision behaviour, producing a corpus of ~800 synthetic transcripts and corresponding audio.

Abstract

Research in emergency triage is restricted to structured electronic health records (EHR) due to regulatory constraints on nurse-patient interactions. We introduce TriageSim, a simulation framework for generating persona-conditioned triage conversations from structured records. TriageSim enables multi-turn nurse-patient interactions with explicit control over disfluency and decision behaviour, producing a corpus of ~800 synthetic transcripts and corresponding audio. We use a combination of automated analysis for linguistic, behavioural and acoustic fidelity alongside manual evaluation for medical fidelity using a random subset of 50 conversations. The utility of the generated corpus is examined via conversational triage classification. We observe modest agreement for acuity levels across three modalities: generated synthetic text, ASR transcripts, and direct audio inputs. The code, persona schemata and triage policy prompts for TriageSim will be available upon acceptance.

TriageSim: A Conversational Emergency Triage Simulation Framework from Structured Electronic Health Records

TL;DR

TriageSim is introduced, a simulation framework for generating persona-conditioned triage conversations from structured records that enables multi-turn nurse-patient interactions with explicit control over disfluency and decision behaviour, producing a corpus of ~800 synthetic transcripts and corresponding audio.

Abstract

Research in emergency triage is restricted to structured electronic health records (EHR) due to regulatory constraints on nurse-patient interactions. We introduce TriageSim, a simulation framework for generating persona-conditioned triage conversations from structured records. TriageSim enables multi-turn nurse-patient interactions with explicit control over disfluency and decision behaviour, producing a corpus of ~800 synthetic transcripts and corresponding audio. We use a combination of automated analysis for linguistic, behavioural and acoustic fidelity alongside manual evaluation for medical fidelity using a random subset of 50 conversations. The utility of the generated corpus is examined via conversational triage classification. We observe modest agreement for acuity levels across three modalities: generated synthetic text, ASR transcripts, and direct audio inputs. The code, persona schemata and triage policy prompts for TriageSim will be available upon acceptance.
Paper Structure (5 sections, 3 figures, 2 tables)

This paper contains 5 sections, 3 figures, 2 tables.

Table of Contents

  1. Introduction
  2. TriageSim Architecture
  3. Quality Evaluation
  4. Triage Classification
  5. Conclusion

Figures (3)

  • Figure 1: Architecture of TriageSim: (a) Input data is constructed using the structured seed data and persona generation. (b) A multi-agent dialogue simulation generates turn-by-turn interaction. (c) Utterances are annotated with phrase structure. (d) Speech is synthesised via accent-conditioned voice cloning and mixed with controlled ambient noise.
  • Figure 2: Evaluation of patient disfluency controllability.
  • Figure 3: Evaluation of nurse behavioural controllability. Left: Effect of experience level on triage confidence. Center: Effect of risk tolerance on over-triage rate. Right: Effect of guideline adherence on vital-sign checking frequency. Error bars denote 95% confidence intervals.