Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Label-dependent and event-guided interpretable disease risk prediction using EHRs

Shuai Niu, Yunya Song, Qing Yin, Yike Guo, Xian Yang

TL;DR

A label-dependent and event-guided risk prediction model (LERP) is proposed to predict the presence of multiple disease risks by mainly extracting information from unstructured medical notes to demonstrate the applicability of the proposed method to the MIMIC-III dataset.

Abstract

Electronic health records (EHRs) contain patients' heterogeneous data that are collected from medical providers involved in the patient's care, including medical notes, clinical events, laboratory test results, symptoms, and diagnoses. In the field of modern healthcare, predicting whether patients would experience any risks based on their EHRs has emerged as a promising research area, in which artificial intelligence (AI) plays a key role. To make AI models practically applicable, it is required that the prediction results should be both accurate and interpretable. To achieve this goal, this paper proposed a label-dependent and event-guided risk prediction model (LERP) to predict the presence of multiple disease risks by mainly extracting information from unstructured medical notes. Our model is featured in the following aspects. First, we adopt a label-dependent mechanism that gives greater attention to words from medical notes that are semantically similar to the names of risk labels. Secondly, as the clinical events (e.g., treatments and drugs) can also indicate the health status of patients, our model utilizes the information from events and uses them to generate an event-guided representation of medical notes. Thirdly, both label-dependent and event-guided representations are integrated to make a robust prediction, in which the interpretability is enabled by the attention weights over words from medical notes. To demonstrate the applicability of the proposed method, we apply it to the MIMIC-III dataset, which contains real-world EHRs collected from hospitals. Our method is evaluated in both quantitative and qualitative ways.

Label-dependent and event-guided interpretable disease risk prediction using EHRs

TL;DR

A label-dependent and event-guided risk prediction model (LERP) is proposed to predict the presence of multiple disease risks by mainly extracting information from unstructured medical notes to demonstrate the applicability of the proposed method to the MIMIC-III dataset.

Abstract

Electronic health records (EHRs) contain patients' heterogeneous data that are collected from medical providers involved in the patient's care, including medical notes, clinical events, laboratory test results, symptoms, and diagnoses. In the field of modern healthcare, predicting whether patients would experience any risks based on their EHRs has emerged as a promising research area, in which artificial intelligence (AI) plays a key role. To make AI models practically applicable, it is required that the prediction results should be both accurate and interpretable. To achieve this goal, this paper proposed a label-dependent and event-guided risk prediction model (LERP) to predict the presence of multiple disease risks by mainly extracting information from unstructured medical notes. Our model is featured in the following aspects. First, we adopt a label-dependent mechanism that gives greater attention to words from medical notes that are semantically similar to the names of risk labels. Secondly, as the clinical events (e.g., treatments and drugs) can also indicate the health status of patients, our model utilizes the information from events and uses them to generate an event-guided representation of medical notes. Thirdly, both label-dependent and event-guided representations are integrated to make a robust prediction, in which the interpretability is enabled by the attention weights over words from medical notes. To demonstrate the applicability of the proposed method, we apply it to the MIMIC-III dataset, which contains real-world EHRs collected from hospitals. Our method is evaluated in both quantitative and qualitative ways.
Paper Structure (17 sections, 5 equations, 2 figures, 1 table)

This paper contains 17 sections, 5 equations, 2 figures, 1 table.

Table of Contents

  1. Introduction
  2. Related Work
  3. Label-dependent Predictive Modelling
  4. Using Clinical Events for Disease Prediction
  5. Methods
  6. Problem Definition and Notations
  7. Model Overview
  8. Embedding Layer
  9. Cross-attention Layer
  10. Fusion and output Layers
  11. Model Training
  12. Experiments
  13. Experimental Dataset
  14. Comparative Methods and Implementation Details
  15. Quantitative analysis
  16. ...and 2 more sections

Figures (2)

  • Figure 1: The structure of the LERP Model. It takes the information from medical notes, clinical events and names of disease risk labels as the inputs. LERP is composed of embedding layers for textual information embedding, cross-attention layer for learning weighted representations of the medical note, and the fusion layer together with the output layer to predict the presence of different disease risks.
  • Figure 2: Case studies to compare the interpretable results from LERP and LERP$^-$. The colour map on the top of this figure maps the colours to normalized attention scores (ranging from 0% to 100%). In the result table, the second/third column contains the clinical events/disease risks associated with the selected fragments of medical notes.