Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

CHiRPE: A Step Towards Real-World Clinical NLP with Clinician-Oriented Model Explanations

Stephanie Fong, Zimu Wang, Guilherme C. Oliveira, Xiangyu Zhao, Yiwen Jiang, Jiahe Liu, Beau-Luke Colton, Scott Woods, Martha E. Shenton, Barnaby Nelson, Zongyuan Ge, Dominic Dwyer

TL;DR

CHiRPE tackles the need for interpretable real-world clinical NLP by processing transcribed PSYCHS interviews to predict CHR-P and generate clinician-informed SHAP explanations. The pipeline combines symptom-domain segmentation, third-person segment summarisation, and transformer-based CHR-P classification, plus five SHAP formats (three co-developed with clinicians) to align explanations with clinical reasoning. Empirical results show AUCs approximately $0.95$–$0.97$ and clear gains from segmentation and clinician-guided explanations, with 28 clinical experts preferring the novel formats over standard SHAP visuals. The work advances practical, interpretable mental health NLP and lays the groundwork for multi-site real-world testing and expansion to additional psychiatric conditions, while acknowledging language and governance limitations and ethical considerations.

Abstract

The medical adoption of NLP tools requires interpretability by end users, yet traditional explainable AI (XAI) methods are misaligned with clinical reasoning and lack clinician input. We introduce CHiRPE (Clinical High-Risk Prediction with Explainability), an NLP pipeline that takes transcribed semi-structured clinical interviews to: (i) predict psychosis risk; and (ii) generate novel SHAP explanation formats co-developed with clinicians. Trained on 944 semi-structured interview transcripts across 24 international clinics of the AMP-SCZ study, the CHiRPE pipeline integrates symptom-domain mapping, LLM summarisation, and BERT classification. CHiRPE achieved over 90% accuracy across three BERT variants and outperformed baseline models. Explanation formats were evaluated by 28 clinical experts who indicated a strong preference for our novel concept-guided explanations, especially hybrid graph-and-text summary formats. CHiRPE demonstrates that clinically-guided model development produces both accurate and interpretable results. Our next step is focused on real-world testing across our 24 international sites.

CHiRPE: A Step Towards Real-World Clinical NLP with Clinician-Oriented Model Explanations

TL;DR

CHiRPE tackles the need for interpretable real-world clinical NLP by processing transcribed PSYCHS interviews to predict CHR-P and generate clinician-informed SHAP explanations. The pipeline combines symptom-domain segmentation, third-person segment summarisation, and transformer-based CHR-P classification, plus five SHAP formats (three co-developed with clinicians) to align explanations with clinical reasoning. Empirical results show AUCs approximately – and clear gains from segmentation and clinician-guided explanations, with 28 clinical experts preferring the novel formats over standard SHAP visuals. The work advances practical, interpretable mental health NLP and lays the groundwork for multi-site real-world testing and expansion to additional psychiatric conditions, while acknowledging language and governance limitations and ethical considerations.

Abstract

The medical adoption of NLP tools requires interpretability by end users, yet traditional explainable AI (XAI) methods are misaligned with clinical reasoning and lack clinician input. We introduce CHiRPE (Clinical High-Risk Prediction with Explainability), an NLP pipeline that takes transcribed semi-structured clinical interviews to: (i) predict psychosis risk; and (ii) generate novel SHAP explanation formats co-developed with clinicians. Trained on 944 semi-structured interview transcripts across 24 international clinics of the AMP-SCZ study, the CHiRPE pipeline integrates symptom-domain mapping, LLM summarisation, and BERT classification. CHiRPE achieved over 90% accuracy across three BERT variants and outperformed baseline models. Explanation formats were evaluated by 28 clinical experts who indicated a strong preference for our novel concept-guided explanations, especially hybrid graph-and-text summary formats. CHiRPE demonstrates that clinically-guided model development produces both accurate and interpretable results. Our next step is focused on real-world testing across our 24 international sites.
Paper Structure (39 sections, 5 figures, 8 tables)

This paper contains 39 sections, 5 figures, 8 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. CHiRPE
  4. Symptom Domain Segmentation
  5. Summarisation of Interview Segments
  6. Classification Models
  7. Explanation Generation with SHAP
  8. Sentence‐level Summary.
  9. Narrative Summary.
  10. Symptom-Level Plots.
  11. Data Analysis Methods
  12. Performance Metrics
  13. Clinical Expert Feedback
  14. Data and Experimental Setup
  15. Dataset
  16. ...and 24 more sections

Figures (5)

  • Figure 1: CHiRPE pipeline. Raw PSYCHS transcripts are segmented by symptom domain and summarised, then passed to a BERT-based classifier. The system outputs a CHR-P or Healthy label alongside SHAP explanations.
  • Figure 2: SHAP word-level bar plot showing the top contributing words for a CHR-P classification decision.
  • Figure 3: Symptom-level plots (red: CHR-P, blue: healthy control).
  • Figure 4: Inline token level heatmaps with SHAP
  • Figure 5: Website Interface.