Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Fairness at Every Intersection: Uncovering and Mitigating Intersectional Biases in Multimodal Clinical Predictions

Resmi Ramachandranpillai, Kishore Sampath, Ayaazuddin Mohammad, Malihe Alikhani

TL;DR

The findings indicate that the proposed sub-group-specific bias mitigation is robust across different datasets, subgroups, and embeddings, demonstrating effectiveness in addressing intersectional biases in multimodal settings.

Abstract

Biases in automated clinical decision-making using Electronic Healthcare Records (EHR) impose significant disparities in patient care and treatment outcomes. Conventional approaches have primarily focused on bias mitigation strategies stemming from single attributes, overlooking intersectional subgroups -- groups formed across various demographic intersections (such as race, gender, ethnicity, etc.). Rendering single-attribute mitigation strategies to intersectional subgroups becomes statistically irrelevant due to the varying distribution and bias patterns across these subgroups. The multimodal nature of EHR -- data from various sources such as combinations of text, time series, tabular, events, and images -- adds another layer of complexity as the influence on minority groups may fluctuate across modalities. In this paper, we take the initial steps to uncover potential intersectional biases in predictions by sourcing extensive multimodal datasets, MIMIC-Eye1 and MIMIC-IV ED, and propose mitigation at the intersectional subgroup level. We perform and benchmark downstream tasks and bias evaluation on the datasets by learning a unified text representation from multimodal sources, harnessing the enormous capabilities of the pre-trained clinical Language Models (LM), MedBERT, Clinical BERT, and Clinical BioBERT. Our findings indicate that the proposed sub-group-specific bias mitigation is robust across different datasets, subgroups, and embeddings, demonstrating effectiveness in addressing intersectional biases in multimodal settings.

Fairness at Every Intersection: Uncovering and Mitigating Intersectional Biases in Multimodal Clinical Predictions

TL;DR

The findings indicate that the proposed sub-group-specific bias mitigation is robust across different datasets, subgroups, and embeddings, demonstrating effectiveness in addressing intersectional biases in multimodal settings.

Abstract

Biases in automated clinical decision-making using Electronic Healthcare Records (EHR) impose significant disparities in patient care and treatment outcomes. Conventional approaches have primarily focused on bias mitigation strategies stemming from single attributes, overlooking intersectional subgroups -- groups formed across various demographic intersections (such as race, gender, ethnicity, etc.). Rendering single-attribute mitigation strategies to intersectional subgroups becomes statistically irrelevant due to the varying distribution and bias patterns across these subgroups. The multimodal nature of EHR -- data from various sources such as combinations of text, time series, tabular, events, and images -- adds another layer of complexity as the influence on minority groups may fluctuate across modalities. In this paper, we take the initial steps to uncover potential intersectional biases in predictions by sourcing extensive multimodal datasets, MIMIC-Eye1 and MIMIC-IV ED, and propose mitigation at the intersectional subgroup level. We perform and benchmark downstream tasks and bias evaluation on the datasets by learning a unified text representation from multimodal sources, harnessing the enormous capabilities of the pre-trained clinical Language Models (LM), MedBERT, Clinical BERT, and Clinical BioBERT. Our findings indicate that the proposed sub-group-specific bias mitigation is robust across different datasets, subgroups, and embeddings, demonstrating effectiveness in addressing intersectional biases in multimodal settings.

Paper Structure

This paper contains 14 sections, 11 equations, 4 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Background
  3. Algorithmic Fairness
  4. Intersectional Fairness
  5. Methodology
  6. Learning Unified Text Representations from Multimodal Data
  7. Embedding and Prediction
  8. Subgroup-specific Discrimination Aware Ensembling (SDAE)
  9. Experiments
  10. Downstream Tasks
  11. Intersectional Bias Analysis and Mitigation
  12. Related Works
  13. Conclusions and Future Works
  14. Reproducibility Checklist

Figures (4)

  • Figure 1: Intersectional biases in multimodal settings; Bias mitigation focused at single attribute is no longer valid for demographic intersections (case 1). The nature of multimodal data adds another layer of complexity in terms of fairness fluctuations (case 2).
  • Figure 2: The proposed architecture
  • Figure 3: Comparison of WP in multitask setting using MIMIC-Eye
  • Figure 4: Comparison of F1 scores