Abstract Rationale, aims, and objectives The impact of teaching versus nonteaching services on outcomes and resource use in patients with acute exacerbation of chronic obstructive pulmonary disease (AECOPD) is unknown. The aim of the study is to evaluate the impact of an internal medicine teaching service compared to a nonteaching service on outcomes and resource use in patients admitted with AECOPD in a community teaching hospital. Methods A retrospective cohort study of patients admitted for a primary diagnosis of chronic obstructive pulmonary disease exacerbation to Florida Hospital Orlando, a large community teaching hospital, between January 1, 2011, and December 31, 2014. Data were extracted from Premier administrative database. Risk adjusted length of stay (LOS), cost of hospitalization, 30‐day readmissions, and mortality rate were measured. Risk adjustment for outcomes was based on Premier CareScience methodology. Results A total of 1419 patients were included, 306 in the teaching group and 1113 in the nonteaching group. Risk adjusted cost and LOS were significantly lower in the teaching group compared to the nonteaching group (observed/expected cost 0.66 vs 1.06, P < .001) and (observed/expected LOS 0.93 vs 1.69, P < .001), respectively. No significant difference was found between the 2 groups in risk adjusted mortality and readmissions ( P = .48 and .89, respectively). Use of consults was significantly lower in the teaching groups with 73% vs 31% of the patient in the teaching group had no consults compared to the nonteaching group ( P < .001). The teaching service was significantly associated with decreased use of consults after adjustment for other variables (odds ratio, 0.17, 95% CI, 0.15‐0.23, P < .001). Conclusion The teaching service had more favorable outcomes compared to nonteaching services in patients hospitalized for AECOPD. The physician practice model has a major impact on the cost, LOS, and use of consults in patients with AECOPD.
Deploying Large Language Models (LLMs) on edge devices is increasingly important, as it eliminates reliance on network connections, reduces expensive API calls, and enhances user privacy. However, on-device deployment is challenging due to the limited computational resources of edge devices. In particular, the key bottleneck stems from memory bandwidth constraints related to weight loading. Weight-only quantization effectively reduces memory access, yet often induces significant accuracy degradation. Recent efforts to incorporate sub-branches have shown promise for mitigating quantization errors, but these methods either lack robust optimization strategies or rely on suboptimal objectives. To address these gaps, we propose FeedBack Quantization (FBQuant), a novel approach inspired by negative feedback mechanisms in automatic control. FBQuant inherently ensures that the reconstructed weights remain bounded by the quantization process, thereby reducing the risk of overfitting. To further offset the additional latency introduced by sub-branches, we develop an efficient CUDA kernel that decreases 60\% of extra inference time. Comprehensive experiments demonstrate the efficiency and effectiveness of FBQuant across various LLMs. Notably, for 3-bit Llama2-7B, FBQuant improves zero-shot accuracy by 1.2\%.
Background: Sleep improvement protocols are recommended for use in the intensive care unit (ICU) despite questions regarding which interventions to include, whether sleep quality or duration will improve, and the role of pharmacists in their development and implementation. Objective: To characterize the impact of a pharmacist-led, ICU sleep improvement protocol on sleep duration and quality as evaluated by a commercially available activity tracker and patient perception. Methods: Critical care pharmacists from a 40-bed, mixed ICU at a large community hospital led the development and implementation of an interprofessional sleep improvement protocol. It included daily pharmacist medication review to reduce use of medications known to disrupt sleep or increase delirium and guideline-based recommendations on both environmental and nonpharmacological sleep-focused interventions. Sleep duration and quality were compared before (December 2018 to December 2019) and after (January to June 2019) protocol implementation in non–mechanically ventilated adults using both objective (total nocturnal sleep time [TST] measured by an activity tracker (Fitbit Charge 2) and subjective (patient-perceived sleep quality using the Richards-Campbell Sleep Questionnaire [RCSQ]) measures. Results: Groups before (n = 48) and after (n = 29) sleep protocol implementation were well matched. After protocol implementation, patients had a longer TST (389 ± 123 vs 310 ± 147 minutes; P = 0.02) and better RCSQ-perceived sleep quality (63 ± 18 vs 42 ± 24 mm; P = 0.0003) compared with before implementation. Conclusion and Relevance: A sleep protocol that incorporated novel elements led to objective and subjective improvements in ICU sleep duration and quality. Application of this study may result in increased utilization of sleep protocols and pharmacist involvement.
The impact of quantization on the overall performance of deep learning models is a well-studied problem. However, understanding and mitigating its effects on a more fine-grained level is still lacking, especially for harder tasks such as object detection with both classification and regression objectives. This work defines the performance for a subset of task-critical categories, i.e. the critical-category performance, as a crucial yet largely overlooked fine-grained objective for detection tasks. We analyze the impact of quantization at the category-level granularity, and propose methods to improve performance for the critical categories. Specifically, we find that certain critical categories have a higher sensitivity to quantization, and are prone to overfitting after quantization-aware training (QAT). To explain this, we provide theoretical and empirical links between their performance gaps and the corresponding loss landscapes with the Fisher information framework. Using this evidence, we apply a Fisher-aware mixed-precision quantization scheme, and a Fisher-trace regularization for the QAT on the critical-category loss landscape. The proposed methods improve critical-category metrics of the quantized transformer-based DETR detectors. They are even more significant in case of larger models and higher number of classes where the overfitting becomes more severe. For example, our methods lead to 10.4% and 14.5% mAP gains for, correspondingly, 4-bit DETR-R50 and Deformable DETR on the most impacted critical classes in the COCO Panoptic dataset.
'/%3e%3cuse xlink:href='%23a' transform='rotate(23.036 .093 25.536)'/%3e%3cuse xlink:href='%23a' transform='rotate(45.87 1.273 16.18)'/%3e%3cuse xlink:href='%23a' transform='rotate(69.945 .996 12.078)'/%3e%3cuse xlink:href='%23a' transform='rotate(20.66 -19.689 31.932)'/%3e%3c/svg%3e)