Category Ranking

98%

Total Visits

921

Avg Visit Duration

2 minutes

Citations

20

Article Abstract

Clinical predictive analysis is a crucial task with numerous applications and has been extensively studied using machine learning approaches. Clinical notes, a vital data source, have been employed to develop natural language processing (NLP) models for risk prediction in healthcare with robust performance. However, clinical notes vary considerably in text composition-written by diverse healthcare providers for different purposes-and the impact of these variations on NLP modeling is also underexplored. It also remains uncertain whether the recent Large Language Models (LLMs) with instruction-following capabilities can effectively handle the risk prediction task out-of-the-box, especially when using routinely collected clinical notes instead of polished text. We address these two important research questions in the context of in-hospital mortality prediction within the critical care setting. Specifically, we propose a supervised hierarchical network with note-specific modules to account for variations across different note categories, and provide a detailed comparison with strong supervised baselines and LLMs. We benchmark 34 instruction-following LLMs based on zero-shot, few-shot, and chain-of-thought prompting with diverse prompt templates. Our results demonstrate that the note-specific network delivers improved risk prediction performance compared to established supervised baselines from both measurement-based and text-based modeling. In contrast, LLMs consistently underperform on this critical task, despite their remarkable performances in other domains. This highlights important limitations and raises caution regarding the use of LLMs for risk assessment in the critical setting. Additionally, we show that the proposed model can be leveraged to select informative clinical notes to enhance the training of other models.

Download full-text PDF

Source
http://dx.doi.org/10.1109/JBHI.2025.3574254DOI Listing

Publication Analysis

Top Keywords

risk prediction
16
clinical notes
16
prediction critical
8
critical care
8
hierarchical network
8
large language
8
language models
8
supervised baselines
8
clinical
6
risk
5

Similar Publications

Systematic analyses uncover plasma proteins linked to incident cardiovascular diseases.

Protein Cell

August 2025

Department of Neurology and National Center for Neurological Disorders, Huashan Hospital, State Key Laboratory of Medical Neurobiology and MOE Frontiers Center for Brain Science, Fudan University, Shanghai 200433, China.

Cardiovascular disease (CVD) research is hindered by limited comprehensive analyses of plasma proteome across disease subtypes. Here, we systematically investigated the associations between plasma proteins and cardiovascular outcomes in 53,026 UK Biobank participants over a 14-year follow-up. Association analyses identified 3,089 significant associations involving 892 unique protein analytes across 13 CVD outcomes.

View Article and Find Full Text PDF

Objectives: The risk of major venous thromboembolism (VTE) among patients with COVID-19 is high but varies with disease severity. Estimate the incidence of lower extremity deep venous thrombosis (DVT) in critically ill hospitalized patients with COVID-19, validate the Wells score for DVT diagnosis, and determine patients' prognosis.

Methods: This was an observational follow-up study in the context of the diagnosis and prognosis of DVT.

View Article and Find Full Text PDF

Objective: To develop and validate a nomogram model for predicting the risk of hyperuricemia (HUA) in perimenopausal women.

Methods: In this study, physical examination information of perimenopausal women was collected at the First Affiliated Hospital of University of Science and Technology of China. We utilized the Least Absolute Shrinkage and Selection Operator (Lasso) and binary logistic regression to investigate the risk factors of HUA among perimenopausal women.

View Article and Find Full Text PDF

Background: Addictive disorders remain a global problem, affecting health, society and the economy. The etiopathogenesis of addictions, which have a multifactorial nature, is poorly understood, making it difficult to develop personalized treatment approaches. Of particular interest is the gene, which regulates serotonergic transmission.

View Article and Find Full Text PDF

Objective: Anoikis is an anchorage-dependent programmed cell death implicated in multiple pathological processes of cancers; however, the prognostic value of anoikis-related genes (ANRGs) in hepatocellular carcinoma (HCC) remains unclear. Our study aims to develop an ANRGs-based prediction model to improve prognostic assessment in HCC patients.

Methods: The RNA-seq profile was performed to estimate the expression of ANRGs in HCC patients.

View Article and Find Full Text PDF