Classification Of Events Research Articles

Multi-scale convolution networks for seismic event classification with windowed self-attention

Multi-scale convolution networks for seismic event classification with windowed self-attention

Real-world research on beta-blocker usage trends in China and safety exploration based on the FDA Adverse Event Reporting System (FAERS).

Beta-blockers are widely used, with continuously updated clinical recommendations. However, their application faces challenges in personalized treatment and safety. The study aimed to investigate the frequency and patterns of prescribing beta-blockers in China and to explore potential adverse event risk signals associated with beta-blockers, providing reference for rational medication use in clinical settings. Prescription data for beta-blockers from January 2018 to June 2023 were extracted through the Hospital Prescription Analysis Collaborative Project in China to analyze clinical usage trends. While adverse drug reaction reports for beta-blockers were obtained from the FDA Adverse Event Reporting System (FAERS) database. The classification and standardization of adverse drug event (ADE) reports were based on the preferred terms (PT) and corresponding system organ classes (SOC) from the Medical Dictionary for Regulatory Activities (MedDRA). Signal detection utilized a proportion imbalance method. In clinical practice, metoprolol dominated beta-blocker prescriptions in China, accounting for 62.2%. Beta-blockers were primarily prescribed to the elderly (65.7%) and male patients (57.0%). However, off-label use of beta-blockers was relatively widespread. For instance, sotalol was prescribed for hypertension at 18.25%, while esmolol was used for angina and heart failure at rates of 12.94% and 14.98%, respectively. In addition, we identified newly discovered adverse reactions associated with beta-blockers, such as BRASH syndrome (metoprolol: n = 186, ROR = 391.285; carvedilol: n = 72, ROR = 256.459), acute kidney injury (bisoprolol: n = 247, ROR = 5.641), premature baby (labetalol: n = 110, ROR = 91.385), and sleep disorder (propranolol: n = 254, ROR = 10.98). Metoprolol led the beta-blocker market in China. Attention was warranted regarding the newly discovered adverse reactions, such as the risk of acute kidney injury with bisoprolol and the potential for BRASH syndrome with metoprolol and carvedilol.

Abstract 4122284: HEART Score Classification of Major Adverse Cardiovascular Events in Patients with Diabetes Mellitus

Introduction: Risk stratifying diabetes mellitus (DM) patients with chest pain in the emergency department (ED) can pose significant challenges because of atypical symptoms, non-specific ECG changes, and elevated baseline troponin levels. This study aims to evaluate HEART Score performance in DM versus non-DM patients. Methods: A prospective observational cohort study included chest pain patients transported by Emergency Medical Services to University of Pittsburgh Medical Center-affiliated hospitals. Data collection involved extracting data from electronic health records, including both DM and non-DM patients with chest pain. The HEART Score was retrospectively calculated by impartial assessors.The primary endpoint was major adverse cardiac events (MACE) within 30 days. This analysis included the classification performance of bivariate comparisons between the HEART Score and ACS outcomes using the area under the receiver-operator characteristic curve (AUC). Results: 4,058 patients were included. 52.7% were male, 42% were white, there was a mean age of 59.24 years (16.13 SD), 28.63% had DM, and 22.1% had MACE. MACE prevalence was 19.7% in non-DM patients and 28% in patients with DM. The AUC demonstrated more robust discriminatory ability of the HEART Score in non-DM (AUC = 0.81) (95% CI 0.79-0.83) versus DM patients (AUC = 0.74) (95% CI 0.71-0.78) (p 0.001). The performance of HEART Score components, including history, age, and risk factors, exhibited better significant predictive capability for MACE in non-DM patients compared to DM patients. Conversely, the HEART-ECG and troponin components were not different between the DM and non-DM groups. The HEART Score demonstrates better performance in White individuals and females when contrasted with males and Black individuals. Nevertheless, these disparities do not reach statistical significance. Conclusion: The HEART Score performs differently in DM versus non-DM patients, with stronger discrimination observed in non-DM individuals. The troponin and ECG components exhibit poorer performance in diabetes mellitus (DM) patients within the context of the HEART Score. This underscores the complexity of risk assessment in DM patients with chest pain and suggests the necessity for further exploration and refinement of risk stratification strategies in this population.

Inequities in Inpatient Pediatric Patient Safety Events by Category.

Few studies have analyzed potential inequities in both pediatric patient safety events (PSEs) and adverse events (AEs) - PSEs leading to harm - nor in PSEs by event type. We sought to examine potential inequities in rates of pediatric PSEs overall, by severity, and by category based on race and ethnicity, insurance payor, and language as measured using voluntary incident reports (IRs). We conducted a retrospective cohort study of pediatric hospitalizations between January 19, 2012 through December 31, 2019 at a US urban, tertiary care children's hospital. Analyzing 85 458 hospitalizations, we compared PSEs overall, by severity, and by event category by race and ethnicity, insurance payor, and language using incident rate ratios (IRRs). In models controlling for covariates, we found that hospitalizations of Latinx (IRR 1.17, 95% confidence interval [CI] 1.07-1.29), non-Latinx Black/African American (IRR 1.17, 95% CI 1.01-1.34), publicly insured (IRR 1.10, 95% CI 1.02-1.20), and nonprivately/nonpublicly insured (IRR 1.12, 95% CI 1.02-1.23) children had higher rates of PSEs compared with reference groups, but the association between language and PSEs was not significant. There were similar patterns among AEs, although only the association between hospitalizations of Latinx patients and AEs was significant. Medication, fluid, or blood and lines or tubes PSEs drove many inequities. We found inequities in PSEs as recorded by IRs, suggesting differences in care related to race, ethnicity, and payor. Limitations include analysis of a single center, that event categories are unique to the institution analyzed, and the voluntary nature of IRs.

Audio-visual event localization with dual temporal-aware scene understanding and image-text knowledge bridging

Audio-visual event localization (AVEL) task aims to judge and classify an audible and visible event. Existing methods devote to this goal by transferring pre-trained knowledge as well as understanding temporal dependencies and cross-modal correlations of the audio-visual scene. However, most works comprehend the audio-visual scene from an entangled temporal-aware perspective, ignoring the learning of temporal dependency and cross-modal correlation in both forward and backward temporal-aware views. Recently, transferring the pre-trained knowledge from Contrastive Language-Image Pre-training model (CLIP) has shown remarkable results across various tasks. Nevertheless, since audio-visual knowledge of the AVEL task and image-text alignment knowledge of the CLIP exist heterogeneous gap, how to transfer the image-text alignment knowledge of CLIP into AVEL field has barely been investigated. To address these challenges, a novel Dual Temporal-aware scene understanding and image-text Knowledge Bridging (DTKB) model is proposed in this paper. DTKB consists of forward and backward temporal-aware scene understanding streams, in which temporal dependencies and cross-modal correlations are explicitly captured from dual temporal-aware perspectives. Consequently, DTKB can achieve fine-grained scene understanding for event localization. Additionally, a knowledge bridging (KB) module is proposed to simultaneously transfer the image-text representation and alignment knowledge of CLIP to AVEL task. This module regulates the ratio between audio-visual fusion features and CLIP’s visual features, thereby bridging the image-text alignment knowledge of CLIP and the audio-visual new knowledge for event category prediction. Besides, the KB module is compatible with previous models. Extensive experimental results demonstrate that DTKB significantly outperforms the state-of-the-arts models.

CNNCat: categorizing high-energy photons in a Compton/Pair telescope with convolutional neural networks

A Compton/Pair telescope, designed to provide spectral resolved images of cosmic photons from sub-MeV to GeV energies, records a wealth of data in a combination of tracking detector and calorimeter. Onboard event classification can be required to decide on which data to down-link with priority, given limited data-transfer bandwidth. Event classification is also the first and one of the most crucial steps in reconstructing data. Its outcome determines the further handling of the event, i.e., the type of reconstruction (Compton, pair) or, possibly, the decision to discard it. Errors at this stage result in misreconstruction and loss of source information. We present a classification algorithm driven by a Convolutional Neural Network. It provides classification of the type of electromagnetic interaction, based solely on low-level detector data. We introduce the task, describe the architecture and the dataset used, and present the performance of this method in the context of the proposed (e-)ASTROGAM and similar telescopes.

China is suffering from fewer but more severe Drought to flood abrupt alternation events

Drought to flood abrupt alternation (DFAA) events, as a special category of compound extreme events that suddenly shift from drought to flood conditions, have significantly greater impacts than individual drought or flood events. In this paper, we have utilized a multifactorial drought index and flood index to identify daily DFAA events occurring in mainland China and in major impact areas during the period 1961-2022. Based on drought and flood index, we have used entropy weighting method to measure the intensity of DFAA events. Our findings indicate that China's DFAA events primarily occur in the hotspots of Huang-Huai-Hai River Basin, the middle and lower Yangtze River Basin, the southeastern coastal area, and the southwestern part of the country. The most frequent and intense DFAA events occur from June to September, with varying subseasonal patterns in the frequency and intensity of events in each hotspot. The frequency of DFAA events in mainland China shows a significant decreasing trend declining at a rate of 0.25 pre year in year-round. While DFAA events occurring in the warm season tend to decrease more significantly compared to the year-round at a rate of 0.26 per year. However, the intensity of DFAA events is increasing with a rate of 0.1 per decade in both the year-round and warm season. The evolution of DFAA events and their direct causes vary non-uniformly across regions and months. Subseasonally, frequency and intensity trends diverged monthly across regions, notably with the Huang-Huai-Hai Basin and southeast coast experiencing a July decline in frequency but a surge in intensity. Our research deepens the understanding of changes in DFAA events and provides practical reference for preventing and mitigating drought-to-flood disasters in mainland China.

Efficacy and safety of low-frequency repetitive transcranial magnetic stimulation in adolescents with first-episode major depressive disorder: A randomized, double-blind, sham-controlled pilot trial

BackgroundA few sham-controlled studies have examined the efficacy, safety, and tolerability of low-frequency repetitive transcranial magnetic stimulation (LF-rTMS) in adolescents with first-episode major depressive disorder (FE-MDD). MethodsForty adolescents (aged 13–17 years) with FE-MDD were randomly assigned to receive active rTMS (n = 20) or sham rTMS (n = 20) for 10 sessions over two weeks. The severity of baseline depressive symptoms and their improvement on the day immediately after the second, fourth, sixth, eighth, and tenth sessions were assessed with the 17-item Hamilton Depression Rating Scale (HAMD-17). Results: After completing 10 rTMS treatment sessions, there was no significant difference in antidepressant response (70.0 % vs. 60.0 %, p > 0.05) and antidepressant remission (55.0 % vs. 35.0 %, p > 0.05) between rTMS groups. The linear mixed model analysis did not show a significant group-by-time interaction (F = 1.26; p > 0.05) in the HAMD-17 scores. There was a significant time main effect on the speed of processing (F = 13.61; p < 0.05), but this did not differ significantly between groups (p > 0.05). There were no other main effects and group-by-time interactions in the other MATRICS Consensus Cognitive Battery domains (all p > 0.05). All adverse event categories, such as fatigue and headache, were similar in the two groups (all p > 0.05). ConclusionsIn this study that compared a combination of LF-rTMS + medication with sham + medication, LF-rTMS had higher response and remission rates than a sham procedure in adolescents with FE-MDD, but the change was not statistically significant. LF-rTMS is generally safe, with mild adverse effects and no negative impact on neurocognitive performance for adolescents with FE-MDD. Registration numberChiCTR2000037878.

Trends in cardiac arrest outcomes & management in children with cardiac illness category compared to non-cardiac illness category: An analysis from the AHA get with the guidelines®-resuscitation registry

IntroductionContemporary rates of survival after pediatric in-hospital CPR events and trends in survival over the last 20 years have not been compared based on illness category. We hypothesized that survival to hospital discharge for surgical-cardiac category is higher than the non-cardiac category, and rates of survival after in-hospital CPR increased over time in all categories. MethodsThe AHA Get With The Guidelines®-Resuscitation registry was queried for index CPR events in children < 18 years of age from 2000 to 2021. Categories were surgical-cardiac (in-hospital CPR event following cardiac surgery); medical-cardiac (CPR event in non-surgical cardiac disease); and non-cardiac (CPR event in patients without cardiac disease). The primary outcome was survival to hospital discharge. We compared eras 2000–2004, 2005–2009, 2010–2014, and 2015–2021 with mixed logistic regression models, including event year as a continuous predictor and site as a random effect. ResultsOf 16,241 index events, in-hospital CPR event rates by illness category were: 19 % surgical-cardiac, 18 % medical-cardiac, and 63 % non-cardiac. Surgical-cardiac category had the highest rate of survival to hospital discharge compared to medical-cardiac and non-cardiac categories (56 % vs. 44 % vs. 46 %; p < 0.001). After controlling for age, location of event, and hospital size, the odds of survival were highest for surgical-cardiac category (aOR 1.28, 95 % CI 1.17–1.41) and lower for medical-cardiac category (aOR 0.90, 0.82–0.98), compared to the non-cardiac category. Odds of survival increased for all illness categories from the 2000–2004 era to the 2015–2021 era. Rates of improvement differed among illness categories with medical-cardiac having the lowest increased odds per era. Surgical-cardiac patients had the highest rates of extracorporeal resuscitation (ECPR) (20 % across the cohort), though the greatest increase in ECPR utilization was in the non-cardiac population (52 % increased odds per era). ConclusionsOver the last 20 years, both survival to hospital discharge and ECPR use has increased in all in-hospital CPR event illness categories. Children with surgical-cardiac CPR event have higher odds of survival to hospital discharge compared to non-cardiac CPR event categories, whereas odds of survival were lowest with medical-cardiac CPR events.

Inpatient Falls: Epidemiology, Risk Assessment, and Prevention Measures. A Narrative Review.

Background Patient falls are a common inpatient dilemma and comprise the largest category of preventable adverse events in hospitalized patients. These events place a clinical burden on the patient, such as increased morbidity and reduced quality of life, in addition to an economic burden on the hospital system. Fall prevention strategies have the opportunity to decrease inpatient health care costs and length of stay. Several risk factors have been identified that contribute to inpatient falls and several strategies have been identified to minimize this risk. These risk factors are typically characterized as intrinsic and extrinsic factors. Intrinsic factors include characteristics such as age, gender, impaired mobility, and physiological factors (eg, co-morbidities, low muscular strength, visual impairment, poor reaction time, and movement disorders). Extrinsic factors are related to characteristics in the environment, such as slippery surfaces, footwear, bad lighting, and the influence of medications. Risk assessment tools, such as the Hendrich II Fall Risk Model, Morse Fall Scale, and STRATIFY, were developed to help identify those at risk. In utilizing these models and our understanding of predisposing risk factors, prevention strategies may be utilized to mitigate these risks. These strategies are often generalized actions including patient education; however, targeted interventions (ie, patient education) also improve outcomes by interrogating specific risk factors.

Hierarchical-Concatenate Fusion TDNN for sound event classification.

Semantic feature combination/parsing issue is one of the key problems in sound event classification for acoustic scene analysis, environmental sound monitoring, and urban soundscape analysis. The input audio signal in the acoustic scene classification is composed of multiple acoustic events, which usually leads to low recognition rate in complex environments. To address this issue, this paper proposes the Hierarchical-Concatenate Fusion(HCF)-TDNN model by adding HCF Module to ECAPA-TDNN model for sound event classification. In the HCF module, firstly, the audio signal is converted into two-dimensional time-frequency features for segmentation. Then, the segmented features are convolved one by one for improving the small receptive field in perceiving details. Finally, after the convolution is completed, the two adjacent parts are combined before proceeding with the next convolution for enlarging the receptive field in capturing large targets. Therefore, the improved model further enhances the scalability by emphasizing channel attention and efficient propagation and aggregation of feature information. The proposed model is trained and validated on the Urbansound8K dataset. The experimental results show that the proposed model can achieve the best classification accuracy of 95.83%, which is an approximate improvement of 5% (relatively) over the ECAPA-TDNN model.

B-TBM: A Novel Deep Learning Model with Enhanced Loss Function for HAZOP Risk Classification Using Natural Language Statistical Laws

HAZOP is a paradigm of industrial safety, and the introduction of deep learning-based HAZOP text categorization marks the arrival of an intelligent era of safety analysis. However, existing risk analysis methods have limitations in processing complex texts and extracting deep risk features. To solve this problem, this paper proposes a novel HAZOP risk event classification model based on BERT, BiLSTM, and TextCNN. The complexity of HAZOP text is revealed by introducing statistical laws of natural language, such as Zipf’s law and Heaps’ law, and the outputs of different levels of BERT are further combined linearly to collaborate with BiLSTM and TextCNN to capture long-term dependency and local contextual information for a more accurate classification task. Meanwhile, an improved loss function is proposed to effectively solve the deficiencies of the traditional cross-entropy loss function in the mislabeling process and improve the generalization ability of the model. It is experimentally demonstrated that the accuracy of the model is improved by 3% to 4% compared to the traditional BERT model in the task of severity and possibility classification of HAZOP reports. This study not only improves the accuracy and efficiency of HAZOP risk analysis, but also provides new ideas and methods for the application of natural language processing in industrial safety.

Methodological framework for early detection of dengue outbreaks in Tanzania

Abstract Background Dengue fever is a major public health problem because it is a widespread and rapidly transmitted mosquito-borne viral disease. We developed a framework for early detection of dengue outbreaks in Tanzania, starting with detection of suspected and/or confirmed cases, followed by site-specific risk characterization, both useful for planning and prioritizing outbreak interventions or epidemic preparedness. Methods Our initial focus was on the identification and categorization of indicators that are specifically tailored to the early detection and classification of outbreak events. The framework was then evaluated using (i) available cases with syndromic and laboratory-confirmed disease information from ProMED emails using decision tree analysis, and (ii) available historical dengue epidemic data at the local level from 2019 with 6,795 confirmed cases using negative binomial regression analysis adjusted for month and area. Results The laboratory-confirmed diagnosis (dengue yes or no) was consistent with the results of the suspected case classification algorithm for clinically defined syndromic cases. There was strong evidence of an increase in dengue cases with higher site-specific risk (rate ratio = 2.51 (95% CI = [1.76, 3.58])) when regressing confirmed dengue cases in 2019 as the dependent variable and site-specific risk as the independent variable. Conclusions The rapid dengue outbreak risk assessment developed may be useful in controlling the epidemic in Tanzania. The suspected case classification algorithm can be a very useful tool to better assess whether a potential dengue case is present and thus requires laboratory confirmation. The framework can be used to rapidly predict the risk of dengue outbreaks, which is useful for planning and prioritizing interventions or for epidemic preparedness. Key messages • Framework for a novel rapid risk assessment for early detection of dengue outbreaks in Tanzania. • Framework relevant to epidemic preparedness.

Hypnotic use and the risk of cardiovascular diseases in insomnia patients

Abstract Background Insomnia, commonly treated with hypnotic agents, is an independent risk factor of cardiovascular diseases. Previous research demonstrated the detrimental effect of hypnotics on the prognosis of cardiovascular patients. Purpose We aimed to examine the association between hypnotic agents and cardiovascular outcomes in general individuals with insomnia. Methods In a propensity score matched cohort of UK Biobank (UKB) participants with insomnia, Cox proportional hazard model was used to estimate the association between regular use of hypnotic agents and predetermined cardiovascular outcomes including incident coronary heart diseases (CHD), heart failure (HF), stroke, and cardiovascular death. Inverse probability of treatment weighting and competing risk models were further performed during sensitivity analysis. Drug-target Mendelian randomization (MR) analyses were employed for further evaluation of hypnotics-related genes and cardiovascular disease association. The genome-wide association data of CHD, HF, and stroke were there large-scale genome-wide association analyses. Tissue-specified expression quantitative trait loci (eQTL) data was derived from BrainSeq phase II. Results During a median follow-up of 14.3 years, the matched cohort documented a total of 1,019 CHD cases, 406 HF cases, 285 stroke cases, and 197 cardiovascular deaths. No significant association was detected between Z-meds and CHD, stroke, and cardiovascular mortality. Benzodiazepine use was significantly associated with the increased risk of CHD, HF, and cardiovascular mortality (Figure 1). The inverse probability of treatment weighting and competing risk models didn’t alter the above associations. Moreover, drug-target MR analyses corroborated the safety of Z-meds in the general population regarding cardiovascular health. However, GABRG1—a drug target for benzodiazepines—was associated with heightened risks for heart failure, and ischemic stroke (Figure 2). Conclusions Our findings suggested the heterogeneous associations between different categories of hypnotics and incident cardiovascular events in individuals with insomnia. Z-meds are safe regarding the risk of cardiovascular outcomes in the UKB population with insomnia.Figure 1Figure 2

A real scale application of a novel set of spatial and similarity features for detection and classification of natural seismic sources from Distributed Acoustic Sensing data

Summary Distributed Acoustic Sensing (DAS) turns a fiber optic into a very dense network of equally-distributed seismic sensors. We focused on the high-density sampling of the seismic wavefield, expressed in strain rates, measured by DAS. Classical approaches used to identify seismic signals rely on the recorded features at one station, but it is difficult to include spatial information in case of dense seismic station networks. This work aims at introducing new spatial and similarity features for seismic event classification suitable to analyze DAS observations. We propose a processing chain based on the XGBoost algorithm and the use of specifically designed spatio-temporal and similarity features for the event classification, and Markov Random Field for the spatial clustering. The methodology is designated to be applied on a continuous stream of DAS observations. We tested our processing chain to detect earthquakes and quarry blasts recorded in the region by permanent seismic networks and included in the RENASS catalog. These events are part of a strain-rate seismic survey carried out during a 3 weeks campaign of DAS measurements along à 91 km fiber optic cable deployed in the central Pyrenees mountains (France). Despite the high anthropogenic activities along the fiber optic path, the proposed method succeeded in detecting earthquakes of magnitude &amp;gt;0.4 and quarry blasts of magnitude &amp;gt;1.0 while limiting the number of false alarms. This performance is particularly noteworthy for low-magnitude events, where detection is accomplished despite a lower signal-to-noise ratio compared to traditional seismometers. The methodology opens the door to real time detection and classification of seismic events measured with long-distance fiber optic systems.

A Machine Learning-based framework and open-source software for Non Intrusive Water Monitoring

Recent research highlights the potential of consumption-based feedback for water conservation, emphasizing the need for Non Intrusive Water Monitoring (NIWM). However, existing NIWM studies often rely on small datasets, a pre-selected class of models, and inaccessible software. Here, we introduce PyNIWM, a machine learning-based open-source Python framework for NIWM. PyNIWM enables water end-use classification via (i) data characterization and feature engineering, (ii) water end-use event classification with four machine learning classifiers, and (iii) performance assessment. We demonstrate PyNIWM on a real-world dataset containing around 800,000 labeled end-use events from 762 homes across the USA and Canada. The four PyNIWM classifiers achieve F1 scores above 0.85, indicating high suitability for water end-use classification. However, a tradeoff between accuracy and computational cost exists. Finally, data balancing through oversampling enhances classification of low-represented end-use classes, but does not improve overall classification. We release PyNIWM as an open-source software, aiming for collaborative and reproducible research.

Multisource Event Classification in ϕ-OTDR Through Sample Feature Synthesis

Multisource Event Classification in ϕ-OTDR Through Sample Feature Synthesis

Domain knowledge-powered attention for air traffic management hazardous events classification

The automated classification of hazardous events in Air Traffic Management (ATM) involves concise texts with specialized terminology. The problem is currently addressed by machine learning text classification methods; however, some methods overlook domain knowledge, while others require it excessively. This research aims to identify a process for integrating low-cost, text-based domain knowledge into classification to address the issue of domain shift. To achieve this, the Wide and Deep Bidirectional Encoder Representations from Transformers (WD-BERT) model is proposed, featuring a unique knowledge-powered attention mechanism. WD-BERT's Wide Attention module assesses the compatibility between input text and domain knowledge to determine class probabilities, while its Deep Attention module extracts contextual features guided by domain knowledge. Additionally, a text mining method is employed to extract domain knowledge from ATM regulation texts. The model is trained and evaluated using a dataset derived from the China ATM Hazard Source Database. It achieves a multi-label classification accuracy of 80.24% and an F1-micro score of 0.88, outperforming comparative models with state-of-the-art performance. Breakdown analyses of the model indicate that the carefully designed attention mechanism endows WD-BERT with the ability to integrate domain knowledge, allowing it to excel in contexts involving complex sentences, multiple terms, and domain shifts. The innovation of this study lies in the proposed domain knowledge-powered attention mechanism, which allows simply organized domain knowledge to effectively guide ATM hazardous events classification, alleviating the impact of domain shift. This method can also be applied to other vertical domains without knowledge graphs.

Addressing the second victim phenomenon among community pharmacists and its impact on clinical pharmacy practice: a consensus study.

The second victim phenomenon, denoting the harmful effects of patient safety incidents on healthcare practitioners, remains insufficiently examined within the pharmacy workforce. This study aimed to investigate the second victim phenomenon in community pharmacies, focusing on its triggers, impacts on pharmacists' well-being, and effects on pharmaceutical care and safety. This consensus study with the Nominal Group Technique involved 27 community pharmacists in three equal groups. The final ranks of the statements scored by participants from 5 to 1 were recalculated using the Van Breda method, combining three distinct data sets with higher values for a higher impact on the output evaluated. Statistics were applied to ascertain event distribution and investigate the potential relationships between event categories and outcomes for patients and pharmacists. "Patient-centric anxiety" (6.8) was the top mental health issue, followed by "Personal responsibility and resilience" and "Future concerns and career aspirations" (6.0 each). The dominant support was "Colleague/Peer support" (5.3). The most frequent patient safety incidents were "Inadequate pharmaceutical service" (8.0) and "Wrong drug dispensed" (7.8). Most errors (63%) were dispensing failures, primarily wrong drug dispensed (44.4%). Of these, 50% were near misses, 25.0% caused no harm, and 16.7% had serious consequences. Field notes suggest contributing factors like inadequate supervision, crowding, and storage issues. This study revealed the second victim phenomenon among pharmacists, which potentially stems from breaches in practice standards. The impact on the quality and safety of pharmaceutical care and its influence on pharmacists' well-being should be studied in further studies.

Does one size fit all? Developing an evaluation strategy to assess large language models for patient safety event report analysis.

Collecting and analyzing patient safety event (PSE) reports is a key component to the improvement of patient safety yet report analysis has been challenging. Large language models (LLMs) may support analysis; however, PSE reports tend to be a hybrid of clinical and general language. We propose a data-driven evaluation strategy to assess LLM fit for report analysis. We identify target tokens and sentences from PSE reports and use perplexity to evaluate four LLMs comprehension of the target sentence. LLMs had statistically significantly different perplexity measures in six of seven event categories. Clinical models perform better with clinical narratives, often reported by nurses and physicians. General models perform better with colloquial language and communication themes. For LLMs to support PSE report analysis there must be a good fit between the language model and the nature of the text in reports. A single LLM approach may not be the most useful strategy.