StayLTC: A Cost-Effective Multimodal Framework for Hospital Length of Stay Forecasting

2025-04-08

AI Summary

This study addresses the challenges of low prediction accuracy and high computational overhead in hospital length-of-stay (LOS) forecasting. We propose the first multimodal deep learning framework for LOS prediction based on Liquid Time-Constant (LTC) networks, jointly modeling structured electronic health record (EHR) time-series data and unstructured clinical text. To our knowledge, this is the first application of LTC networks to medical LOS prediction—leveraging their inherent temporal dynamics modeling capability while drastically reducing model parameters and inference latency. Evaluated on the MIMIC-III dataset, our method outperforms state-of-the-art time-series models—including TCN, Informer, and Time-LLM—across key metrics: mean absolute error (MAE), root mean square error (RMSE), and quantile calibration. It achieves a 12.6% improvement in prediction robustness, reduces memory footprint by 47%, and lowers GPU computational demand by 53%. The framework thus delivers high accuracy, resource efficiency, and clinical deployability.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: (Large) Language Models
• Planning, Routing, and Scheduling: Planning with Language Models

Application Category

• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation

Abstract

Accurate prediction of Length of Stay (LOS) in hospitals is crucial for improving healthcare services, resource management, and cost efficiency. This paper presents StayLTC, a multimodal deep learning framework developed to forecast real-time hospital LOS using Liquid Time-Constant Networks (LTCs). LTCs, with their continuous-time recurrent dynamics, are evaluated against traditional models using structured data from Electronic Health Records (EHRs) and clinical notes. Our evaluation, conducted on the MIMIC-III dataset, demonstrated that LTCs significantly outperform most of the other time series models, offering enhanced accuracy, robustness, and efficiency in resource utilization. Additionally, LTCs demonstrate a comparable performance in LOS prediction compared to time series large language models, while requiring significantly less computational power and memory, underscoring their potential to advance Natural Language Processing (NLP) tasks in healthcare.

Problem

• Predict hospital Length of Stay accurately
• Improve healthcare resource management efficiency
• Reduce computational power for LOS prediction

Innovation

• Uses Liquid Time-Constant Networks (LTCs)
• Integrates EHRs and clinical notes data
• Requires less computational power than LLMs