Clinical decisions, such as diagnosing conditions, prescribing medications, and recommending procedures, are rarely made with absolute certainty. Instead, they reflect probabilistic judgments shaped by evolving patient information and incomplete evidence. However, current EHR systems and knowledge graphs encode such decisions as deterministic triples, lacking a mechanism to represent the subjective confidence inherent in clinical reasoning. We present JudgEHR, a framework for clinical decision confidence estimation that leverages large language models (LLMs) to perform cohort-based collective inference over structured patient records by representing clinical events as knowledge graph triples and integrating them into LLM prompts. JudgEHR groups related clinical concepts into cohorts using LLM-driven relational inference, and then jointly evaluates the plausibility of all clinical decision triples within each cohort by considering patient visit history and background medical knowledge. We apply our method to the MIMIC-III dataset. Our statistical analysis shows that JudgEHR generates semantically consistent confidence scores, with similar concepts receiving closer values, whereas dissimilar replacements yield large confidence differences. Experiments on the MIMIC-III dataset show that incorporating the confidence into a zero-shot LLM-based pipeline improves relative AUROC by ${1 4. 6 \%}$ and AUPRC by 21.8% on the mortality prediction task.
Kexuan Xin, Guillaume Pelat, Jonathan Vitale et al.· International Conference on...· 0 citations
Dementia affects over 57 million people worldwide and places an immense burden on informal caregivers, yet current AI tools remain largely fragmented across isolated tasks and modalities. Recent large language models (LLMs) and vision-language models (VLMs) offer promising capabilities for dementia support, but adapting them to this safety-critical, multimodal, and deeply individualized care domain raises challenges that general-purpose AI surveys do not address. In this paper, we present a challenge-driven survey that organizes the rapidly growing literature on LLMs and VLMs for dementia care around three core adaptation challenges: (1) Knowledge Grounding, which anchors model outputs to verified clinical knowledge through retrieval augmented generation, knowledge graphs, and constrained training to mitigate hallucination risk; (2) Multimodal Understanding, which fuses visual, audio, and sensor data with language to reason about the multimodal inherent of daily care; and (3) Personalization, which adapts model behavior to individual patient histories, caregiver needs, and unique disease progression over time via persistent memory and biography-driven interaction. We review over 20 recent methods, identify cross-cutting architectural patterns, and survey available datasets and benchmarks. Finally, we highlight critical open challenges including the need for standardized evaluation protocols, longitudinal deployment studies, and tighter integration between clinical workflows and foundation model capabilities.
Afrouz Sheikholeslami, Dexuan Ding, Amin Beheshti et al.· International Conference on...· 0 citations