BACKGROUND: Large language models (LLMs) have the potential to support promising new applications in health informatics. However, practical data on sample size considerations for fine-tuning LLMs to perform specific tasks in biomedical and health policy contexts are lacking.
OBJECTIVE: This study aims to evaluate sample size and sample selection techniques for fine-tuning LLMs to support improved named entity recognition (NER) for a custom data set of conflicts of interest disclosure statements.
METHODS: A random sample of 200 disclosure statements was prepared for annotation. All \"PERSON\" and \"ORG\" entities were identified by each of the 2 raters, and once appropriate agreement was established, the annotators independently annotated an additional 290 disclosure statements. From the 490 annotated documents, 2500 stratified random samples in different size ranges were drawn. The 2500 training set subsamples were used to fine-tune a selection of language models across 2 model architectures (Bidirectional Encoder Representations from Transformers [BERT] and Generative Pre-trained Transformer [GPT]) for improved NER, and multiple regression was used to assess the relationship between sample size (sentences), entity density (entities per sentence [EPS]), and trained model performance (F1-score). Additionally, single-predictor threshold regression models were used to evaluate the possibility of diminishing marginal returns from increased sample size or entity density.
RESULTS: Fine-tuned models ranged in topline NER performance from F1-score=0.79 to F1-score=0.96 across architectures. Two-predictor multiple linear regression models were statistically significant with multiple R2 ranging from 0.6057 to 0.7896 (all P<.001). EPS and the number of sentences were significant predictors of F1-scores in all cases ( P<.001), except for the GPT-2_large model, where EPS was not a significant predictor (P=.184). Model thresholds indicate points of diminishing marginal return from increased training data set sample size measured by the number of sentences, with point estimates ranging from 439 sentences for RoBERTa_large to 527 sentences for GPT-2_large. Likewise, the threshold regression models indicate a diminishing marginal return for EPS with point estimates between 1.36 and 1.38.
CONCLUSIONS: Relatively modest sample sizes can be used to fine-tune LLMs for NER tasks applied to biomedical text, and training data entity density should representatively approximate entity density in production data. Training data quality and a model architecture\'s intended use (text generation vs text processing or classification) may be as, or more, important as training data volume and model parameter size.

BACKGROUND: E-cigarette use is most prevalent among adolescents and young adults - and there are often misperceptions about product risk. The purpose of this study was to determine what nicotine information is provided on e-cigarette brand websites.
METHODS: Based on national and local surveys, we identified 44 e-cigarette brands commonly used in the US by adolescents and young adults. For each of these brands, their associated websites were analyzed for disclosed nicotine information. Specifically, for each brand\'s website, we coded whether there was information on nicotine concentration (recorded if a numerical value was provided such as \'5% nicotine\'), nicotine form (free-base, nicotine salts, or not stated), and nicotine type (tobacco-derived, synthetically derived, or not stated). Coding allowed for both lay (e.g. \'nic salts\') as well as scientific (e.g. \'isomers\') terms.
RESULTS: Of the 44 brands examined, all provided basic information on nicotine concentration (e.g. \'5% nicotine\'). However, 23% of brands did not disclose information on nicotine form (i.e. nicotine salt vs free-base), and 66% of brands did not disclose information on nicotine type (i.e. synthetic vs tobacco-derived).
CONCLUSIONS: Overall, these results suggest that the e-cigarette industry is not fully informing its consumers about the nicotine in their products. Given that nicotine form and type have implications for e-cigarette addiction potential, these findings highlight a public health concern. There is a need for more comprehensive national regulations for mandating product constituents and emissions disclosures.

BACKGROUND: The accuracy and completeness of self-disclosures by authors of imaging guidelines are not well known.
OBJECTIVE: The aim of this study was to assess the accuracy of financial disclosures by US authors of ACR appropriateness criteria.
METHODS: We reviewed financial disclosures provided by US-based authors of all ACR-AC published in 2019, 2021 and 2023. For each US- based author, payment reports were extracted from the Open Payments Database (OPD) in the previous 36 months related to general category and research payments categories. We analyzed each author individually to determine if the reported disclosures matched results from OPD.
RESULTS: A total of 633 authorships, including 333 unique authors were included from 38 ACR AC articles in 2019, with 606 authorships (387 unique authors) from 35 ACR-AC articles published in 2021, and 540 authorships (367 unique authors) from 32 ACR AC articles published in 2023. Among authors who received industry payments, failure to disclose any financial relationship was seen in 125/147 unique authors in 2019, 142/148 authors in 2021 and 95/125 unique authors in 2023. The proportion of nondisclosed total value of payments was 86.1% in 2019, 88.6% in 2021 and 56.7% in 2023. General category payments were nondisclosed in 94.1% in 2019, 89.7% in 2021 and 94.4% in 2023 by payment value.
CONCLUSIONS: Industry payments to authors of radiology guidelines are common and frequently undisclosed.

暂无摘要。

Dataset in this article contains solid information on the novel dataset of exceptional women directors (EWDs) and carbon information disclosures (CID) of global energy leading companies. The data presented is related to the research article entitled \"The Effect of Women\'s Leadership on Carbon Disclosure by the Top 100 Global Energy Leaders\" [1]. In exploring the degree of EWDs\' and the level of CID, the content analysis technique based on the enhanced scoring indicators were deployed to obtain the data of the 97 companies based on accessible materials from companies\' websites or any associated reports such as sustainability, environmental, annual, or integrated reports within three-year periods (2018 - 2020) for the global energy leading companies which listed in Thomson Reuters Database 2017. Data on EWDs\' engagement and CID are extracted from the available information, reports, and materials. The data are collected based on the enhanced version of the EWD\'s indicators adapted from Ramon Llorens, García-Meca and Pucheta-Martínez [2], Hillman, Cannella and Paetzold [3], and Jaaffar and Amran [4]. Meanwhile, data collected for CID by using Carbon Disclosures Index (CDI) comprising 90 scores within nine aspects; Strategy and Policy; Climate Change Risks and Opportunities; Corporate GHG Emissions Targets; Company Wide Carbon Footprint; GHG Emissions Change Over Time; Energy-Related Reporting; Emission Reduction Initiatives Implementation; Carbon Emission Accountability; and Quality of Disclosure as suggested by the prior scholars [5], [6], [7]. This dataset shed light as an indicator to measure exceptional women director in the energy companies in promoting transparent carbon disclosure performance as well as boosting women leadership involvement and participation in the most polluting sector which aligned with the United Nations Sustainable Development Goals; SDG 5 gender equality, SDG 10 reduced inequality, and SDG 13 climate action.

Suicidal ideation and suicidal behavior (SI/SB) are prevalent among adolescents. Treatment of SI/SB in adolescents relies on their disclosure, yet there is limited research on adolescent SI/SB disclosure experiences. Understanding who they disclose to and how they experience their parents\' responses to their disclosures is important, as parents are often involved in adolescent mental health treatment.
The present study characterized adolescent SI/SB disclosures in a sample of psychiatrically hospitalized adolescents, examining to whom they disclosed SI/SB, perceived parental responses to SI/SB disclosures, and what they would prefer their parents did differently in response to SI/SB disclosures.
Results indicate that over 50% of youth disclosed their SI/SB directly to their parent and approximately 15%-20% of youth did not disclose their SI/SB to anyone prior to psychiatric hospitalization. Perceived parental responses to disclosures varied, including both validating and invalidating responses.
Findings have important implications for supporting parents and adolescents in discussing SI/SB.

The Asia-Pacific region (APAC) represents a unique environment for the publication of biomedical research, particularly industry-funded research. Awareness and adoption of international guidelines on ethical publication practices continues to increase across APAC, but the reframing and expansion of many of the recommendations in the Good Publication Practice (GPP) 2022 guidelines versus GPP3 published in 2015 have important implications for publishing industry-funded biomedical research in the region.
This manuscript provides practical guidance for stakeholders in APAC on interpreting and applying the recommendations made in the GPP 2022 guidelines.
Key focus areas include navigating new opportunities for communicating industry-funded research, such as plain language summaries, social media, and preprints; implementing formal processes to improve the integrity of published research in APAC; and methods of promoting transparency and inclusion when publishing industry-funded research. Key APAC-specific issues, including encore presentations, leadership on publication ethics in the region, access to professional resources, and support for educating regional stakeholders are also discussed.
Overall, this manuscript offers a pragmatic guide for stakeholders in industry-sponsored research on applying GPP 2022 in practice with a focus on effectively integrating these guidelines in an APAC context.

This study examined how people interpret overall survival (OS), overall response rate (ORR), and progression-free survival (PFS) endpoints in the context of direct-to-consumer television ads. Although there is little research on this topic, initial evidence suggests that people can misinterpret these endpoints. We hypothesized that understanding of ORR and PFS would be improved by adding a disclosure (\"We currently do not know if [Drug] helps patients live longer\") to ORR and PFS claims.
We conducted 2 online studies with US adults examining television ads for fictional prescription drugs indicated to treat lung cancer (N = 385) or multiple myeloma (N = 406). The ads included claims about OS, ORR with and without a disclosure, or PFS with and without a disclosure. In each experiment, we randomized participants to view 1 of 5 versions of a television ad. After viewing the ad twice, participants completed a questionnaire that measured understanding, perceptions, and other outcomes.
In both studies, participants correctly differentiated between OS, ORR, and PFS via open-ended responses; however, participants in the PFS conditions (versus ORR conditions) were more likely to make incorrect inferences about OS. Supporting the hypothesis, adding a disclosure made expectations around living longer and quality-of-life improvements more accurate.
Disclosures could help reduce the extent to which people misinterpret endpoints like ORR and PFS. More research is needed to establish best-practice recommendations for using disclosures to improve patient understanding of drug efficacy without changing their perception of the drug in unintended ways.

This review examines data from stone conferences and research journals to assess whether justifiable concerns exist about possible bias in urolithiasis research and if so, how they can be minimized. Conflict of interest (COI) policies of two major urological congresses and three symposia dedicated to stone research were reviewed. Disclosure slides were viewed in webcasts and were evaluated for robustness and speaker compliance with respect to policy. Additionally, disclosure and COI policies of ten Science Citation Index (SCI)-approved journals were assessed and compared with actual declarations in published papers on urolithiasis. It was observed that disclosure and conflict declarations are frequently conflated in congresses and journals. Differences between the two ideologies appear to be ignored or unappreciated. Disclosures in the major urological meetings revealed a high percentage of financial relationships with industry. In dedicated stone conferences, more than two-thirds of speakers failed to display a declaration slide. Both scenarios generate questions about objectivity. Disclosure and COI statements in journals varied widely in format, detail and content. It is concluded that there exists a misinformed and incorrect perception in urolithiasis research that disclosure of potential COIs somehow validates a study as being objective and unbiased. Current policies and practices at conferences and in published papers create a setting in which concerns of bias prevail. Changes, including the establishment of a universal policy, insistence of independent and explicit declarations of disclosures and conflicts, implementation of sanctions for transgression and the introduction of intensive scrutiny by reviewers are required to minimize doubts.

The health sector in Australia and the ASX100 is lagging far behind in the implementation of carbon management and climate risk analysis. This case study highlights the low quantity and quality of the sector compared to its market weight. The analysis of CDP disclosures for Australian healthcare companies shows this delay and a general lack of interest in the Task Force on Climate-Related Financial Disclosures\' (TCFD) recommendations. Yet, the physical and transitory risks for these companies do exist. The reasons for this inaction represent a knowledge gap in the literature, but several hypotheses are formulated, such as the lack of pressure from public authorities. At the level of the ten largest healthcare companies in the world, this failure to act is not systemic, so the scope of analysis must be broadened to see a pattern emerging.