LAVA: Language Model Assisted Verbal Autopsy for Cause-of-Death Determination

2025-09-11

AI Summary

In resource-limited settings, the absence of formal medical certification severely compromises the accuracy of verbal autopsy (VA) in determining cause of death. To address this, we propose an automated VA method integrating large language models (LLMs) with traditional algorithms. Our approach uniquely combines an off-the-shelf LLM (GPT-5), text embedding models, the LCVA baseline, and a meta-learning ensemble strategy into a multimodal VA analytical framework. Evaluated on the PHMRC gold-standard dataset, it achieves mean diagnostic accuracies of 48.6%, 50.5%, and 53.5% for adult, child, and neonatal subgroups, respectively—surpassing conventional methods by 5–10 percentage points. These results demonstrate that an LLM-driven, lightweight ensemble paradigm significantly enhances the practicality and generalizability of VA for public health surveillance in low-resource contexts.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: Language Grounding & Multi-modal NLP
• Knowledge Representation and Reasoning: Diagnosis and Abductive Reasoning

Application Category

• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics

Abstract

Verbal autopsy (VA) is a critical tool for estimating causes of death in resource-limited settings where medical certification is unavailable. This study presents LA-VA, a proof-of-concept pipeline that combines Large Language Models (LLMs) with traditional algorithmic approaches and embedding-based classification for improved cause-of-death prediction. Using the Population Health Metrics Research Consortium (PHMRC) dataset across three age categories (Adult: 7,580; Child: 1,960; Neonate: 2,438), we evaluate multiple approaches: GPT-5 predictions, LCVA baseline, text embeddings, and meta-learner ensembles. Our results demonstrate that GPT-5 achieves the highest individual performance with average test site accuracies of 48.6% (Adult), 50.5% (Child), and 53.5% (Neonate), outperforming traditional statistical machine learning baselines by 5-10%. Our findings suggest that simple off-the-shelf LLM-assisted approaches could substantially improve verbal autopsy accuracy, with important implications for global health surveillance in low-resource settings.

Problem

• Improving verbal autopsy accuracy for cause-of-death determination
• Combining LLMs with traditional methods for death prediction
• Enhancing global health surveillance in low-resource settings

Innovation

• Combining LLMs with traditional algorithms
• Using text embeddings for classification
• Employing meta-learner ensembles approach