Background: The global nursing shortage, exacerbated by heavy workloads and high turnover rates associated with the COVID-19 pandemic, continues to undermine care quality and nurse well-being. Although digital health technologies have enhanced coordination, improved communication, and reduced clinical errors in nursing practice, they have also increased nurses’ documentation burden. Advances in large language models (LLMs) and other generative artificial intelligence (GenAI) tools facilitate the generation of accurate reports from electronic medical records (EMRs), thereby streamlining documentation workflows, saving time, and reducing nurses’ workloads. Accordingly, integrating LLMs into electronic nursing documentation systems warrants further exploration. Objective: This study examines the integration of an LLM into an in-house nursing information system (NIS) implemented across 3 hospitals in Taiwan to reduce the time and effort required for nursing handover documentation and to preliminarily assess the operational and economic implications of GenAI-assisted workflows. Methods: A multidisciplinary team of nursing specialists and information technology experts at Taipei Medical University (TMU) restructured the organization’s existing nursing handover documentation process to facilitate interaction with the LLM. The team also developed prompt-based interfaces to automatically generate section-specific content for the nursing handover document. The LLM-integrated NIS was subsequently deployed across 3 hospitals in Taiwan: Taipei Medical University Hospital (TMUH), Wan Fang Hospital (WFH), and Shuang Ho Hospital (SHH). We then extracted and analyzed NIS log data to compare documentation times before and after LLM implementation, thereby quantifying time savings. Results: Integration of the LLM into nursing handover documentation was associated with shorter per-patient documentation time in routine clinical use across TMUH, WFH, and SHH. Based on preintegration NIS logs (September 2024), the average handover document completion time per patient ranged from 3.45 (SD 3.82) to 4.32 (SD 4.48) minutes across hospitals and shifts, providing a preliminary baseline for subsequent comparisons. In postintegration NIS logs (October-December 2024), the overall handover document completion time per patient (mean) was substantially lower, ranging from 1.17 (SD 1.86) to 2.54 (SD 2.82) minutes across hospitals and shifts. Using monthly patient volume to estimate time savings, 113-273, 160-314, and 198-391 hours were saved per month at TMUH, WFH, and SHH, respectively, corresponding to aggregate savings of 474-981 hours per month across hospitals during the study period. Conclusions: We integrated an LLM into an NIS to generate nursing handover documents without altering existing workflows. Across 3 hospitals within TMU’s health system, GenAI assistance was associated with shorter documentation time and a positive net labor value from October to December 2024. Prompts were constrained, and nurse verification was required to mitigate hallucinations. Future work will enhance logging to capture reliability and editing metrics, compare LLM-generated drafts with nurse-finalized notes to inform prompt refinement, and assess generalizability to other documentation workflows.

arXiv:2603.02123v2 Announce Type: replace Abstract: The development of affective multimodal language models (MLMs) has long been constrained by a gap between low-level perception and high-level interaction, leading to fragmented affective capabilities and limited generalization. To bridge this gap, we propose a cognitively inspired three-level hierarchy that organizes affective tasks according to their cognitive depth-perception, understanding, and interaction-and provides a unified conceptual foundation for advancing affective modeling. Guided by this hierarchy, we introduce Nano-EmoX, a small-scale multitask MLM, and P2E (Perception-to-Empathy), a curriculum-based training framework. Nano-EmoX integrates a suite of omni-modal encoders, including an enhanced facial encoder and a fusion encoder, to capture key multimodal affective cues and improve cross-task transferability. The outputs are projected into a unified language space via heterogeneous adapters, empowering a lightweight language model to tackle diverse affective tasks. Concurrently, P2E progressively cultivates emotional intelligence by aligning rapid perception with chain-of-thought-driven empathy. To the best of our knowledge, Nano-EmoX is the first compact MLM (2.2B) to unify six core affective tasks across all three hierarchy levels, achieving state-of-the-art or highly competitive performance across multiple benchmarks, demonstrating excellent efficiency and generalization.