Can we generate portable representations for clinical time series data using LLMs?

2026-03-25

AI Summary

This work addresses the significant performance degradation of clinical machine learning models when deployed across hospitals, primarily caused by distribution shifts stemming from a lack of transferable patient representations. To overcome this, the authors propose a lightweight, no-fine-tuning-required portable representation method: irregular ICU time series are converted into natural language summaries via structured prompts using a frozen large language model (LLM), and these summaries are then encoded into fixed-length patient vectors by a frozen text embedding model for downstream prediction tasks. Evaluated on three cohorts—MIMIC-IV, HiRID, and PPICU—the approach matches state-of-the-art baselines in within-hospital performance while exhibiting smaller performance drops, lower prediction variance, and reduced leakage of sensitive attributes during cross-center deployment. It also substantially lowers engineering overhead and enhances few-shot learning efficacy.

Technology Category

• Machine Learning: Representation Learning
• Natural Language Processing: (Large) Language Models
• Computer Vision: Representation Learning for Vision

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Deploying clinical ML is slow and brittle: models that work at one hospital often degrade under distribution shifts at the next. In this work, we study a simple question -- can large language models (LLMs) create portable patient embeddings i.e. representations of patients enable a downstream predictor built on one hospital to be used elsewhere with minimal-to-no retraining and fine-tuning. To do so, we map from irregular ICU time series onto concise natural language summaries using a frozen LLM, then embed each summary with a frozen text embedding model to obtain a fixed length vector capable of serving as input to a variety of downstream predictors. Across three cohorts (MIMIC-IV, HIRID, PPICU), on multiple clinically grounded forecasting and classification tasks, we find that our approach is simple, easy to use and competitive with in-distribution with grid imputation, self-supervised representation learning, and time series foundation models, while exhibiting smaller relative performance drops when transferring to new hospitals. We study the variation in performance across prompt design, with structured prompts being crucial to reducing the variance of the predictive models without altering mean accuracy. We find that using these portable representations improves few-shot learning and does not increase demographic recoverability of age or sex relative to baselines, suggesting little additional privacy risk. Our work points to the potential that LLMs hold as tools to enable the scalable deployment of production grade predictive models by reducing the engineering overhead.

Problem

• portable representations
• clinical time series
• distribution shift
• model transferability
• LLMs

Innovation

• portable patient embeddings
• large language models
• clinical time series
• frozen LLM
• cross-hospital generalization