Explore the vast range of titles available.

Misleading or unnecessary data can have out-sized impacts on the health or accuracy of Machine Learning (ML) models. We present a Bayesian sequential selection method, akin to Bayesian experimental design, that identifies critically important information within a dataset while ignoring data that are either misleading or bring unnecessary complexity to the surrogate model of choice. Our method improves sample-wise error convergence and eliminates instances where more data lead to worse performance and instabilities of the surrogate model, often termed sample-wise “double descent”. We find these instabilities are a result of the complexity of the underlying map and are linked to extreme events and heavy tails. Our approach has two key features. First, the selection algorithm dynamically couples the chosen model and data. Data is chosen based on its merits towards improving the selected model, rather than being compared strictly against other data. Second, a natural convergence of the method removes the need for dividing the data into training, testing, and validation sets. Instead, the selection metric inherently assesses testing and validation error through global statistics of the model. This ensures that key information is never wasted in testing or validation. The method is applied using both Gaussian process regression and deep neural network surrogate models.

Various methods are employed to recover patterns of area relationships in extinct and extant clades. The fidelity of these patterns can be adversely affected by sampling error in the form of missing data. Here we use simulation studies to evaluate the sensitivity of an analytical biogeographical method, namely tree reconciliation analysis (TRA), to this form of sampling failure. Simulation study. To approximate varying degrees of taxonomic sampling failure within phylogenies varying in size and in redundancy of biogeographical signal, we applied sequential pruning protocols to artificial taxon-area cladograms displaying congruent patterns of area relationships. Initial trials assumed equal probability of sampling failure among all areas. Additional trials assigned weighted probabilities to each of the areas in order to explore the effects of uneven geographical sampling. Pruned taxon-area cladograms were then analysed with TRA to determine if the optimal area cladograms recovered match the original biogeographical signal, or if they represent false, ambiguous or uninformative signals. The results indicate a period of consistently accurate recovery of the true biogeographical signal, followed by a nonlinear decrease in signal recovery as more taxa are pruned. At high levels of sampling failure, false biogeographical signals are more likely to be recovered than the true signal. However, randomization testing for statistical significance greatly decreases the chance of accepting false signals. The primary inflection of the signal recovery curve, and its steepness and slope depend upon taxon-area cladogram size and area redundancy, as well as on the evenness of sampling. Uneven sampling across geographical areas is found to have serious deleterious effects on TRA, with the accuracy of recovery of biogeographical signal varying by an order of magnitude or more across different sampling regimes. These simulations reiterate the importance of taxon sampling in biogeographical analysis, and attest to the importance of considering geographical, as well as overall, sampling failure when interpreting the robustness of biogeographical signals. In addition to randomization testing for significance, we suggest the use of randomized sequential taxon deletions and the construction of signal decay curves as a means to assess the robustness of biogeographical signals for empirical data sets.

The risky choice framing effect disclosed that presenting data in a loss scenario lead decision-makers towards risky choices. Conversely, a gain scenario prevents them from taking a risk. Framing effect robustness has been widely confirmed by psychological literature. However, the framing of medical treatments, based on McNeil et al. (1982) paradigm, raised both methodological doubts and contrasting evidence. Our research aimed to investigate the presence and the nature of the framing effect in the McNeil et al. (1982) paradigm. In particular, we thought that the obtained switch of preferences across frames was due to a misleading formulation of the data given in a negative cumulative frequency format. We conducted three studies: (1) we replicated McNeil et al.’s (1982) original study (N=150) with medicine (n=50), statistics (n=50) and lay (n=50) students; (2) we tested (N=180) our hypothesis by comparing a cumulative frequency format with an alternative version, namely a linear progression one; (3) we compared (N=430) the effect of different formats (cumulative frequency, linear progression and interval frequency) on choices. Our results showed that, while the framing effect is present when employing a cumulative frequency format, it disappears when using a linear progression one. Moreover, our results show that decision-makers better understand information when given in a linear progression and an interval frequency format. In the current paper, we argue that the way in which a problem is formulated plays a relevant role in the representation of the decisional task and the decision-making. Keywords: medical framing effect, reverse pattern of choice, understanding numerical information, pragmatic approach.

Electronic health record (EHR) data are increasingly used in predictive models of posttraumatic stress disorder (PTSD), but it is unknown how multivariable prediction of an EHR-based diagnosis might differ from prediction of a more rigorous diagnostic criterion. This distinction is important because EHR data are subject to multiple biases, including diagnostic misclassification and differential health care use resulting from factors such as illness severity. This study aims to compare predictive models using the same predictors to predict an EHR-based versus semistructured interview-based PTSD diagnostic criterion, quantify model performance discrepancies, and examine potential mechanisms that account for performance differences. We compared the performance of several machine learning models predicting EHR-based PTSD diagnosis to models predicting semistructured interview-based diagnosis in a nationwide sample of 1343 US veterans who completed Structured Clinical Interview for DSM-5 (Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition) (SCID-5) interviews and had clinic visit data extracted from the Veterans Affairs (VA) EHR. We developed 2 sets of predictive models using 3 algorithms (elastic net regression, random forest, and XGBoost), with a nested cross-validation scheme consisting of an initial train-test split and 10-fold cross-validation within the training set for each type of model. All models used a nearly identical set of predictors including 29 EHR-based visit count variables and 8 demographic variables. Diagnostic concordance between EHR-based PTSD diagnosis and SCID-5-based PTSD diagnosis was 73.3%, with 17.8% false negatives and 8.9% false positives for EHR-based diagnosis. Models predicting EHR-based PTSD performed very well (area under the receiver operating characteristic curve [AUC] .85-.9; Matthews correlation coefficient [MCC] .58-.69), whereas those predicting interview-based PTSD performed only moderately well overall (AUC .71-.76; MCC .24-.28). Sensitivity analyses showed that participants' frequency of VA visits played a role in these differences, such that the density of EHR data (proportion of nonzero visit counts across EHR variables) was more associated with EHR-based PTSD diagnosis (b=-0.18, SE 0.02, P<.001) than with SCID-5 interview-based PTSD diagnosis (b=-0.06, SE 0.01, P<.001). Predictive models of PTSD built using only EHR data demonstrated inflated performance metrics relative to models predicting diagnosis from a rigorous structured clinical interview. This performance discrepancy appears driven by circular relationships between health care use patterns and EHR-based diagnosis that do not affect external diagnostic criteria. Researchers building clinical prediction models should not assume that diagnosis in the EHR is a sufficient proxy for the true criterion of interest. Clinicians and researchers should be cautious in interpreting clinical prediction models using only EHR data, as their real-world utility may be less than performance metrics suggest.

Misinformation poses a serious challenge to clinical and policy decision-making in the health field. The COVID-19 pandemic amplified interest in misinformation and related terms and witnessed a proliferation of definitions. We aim to assess the definitions of misinformation and related terms used in health-related literature. We conducted a scoping review of systematic reviews by searching Ovid MEDLINE, Embase, Cochrane, and Epistemonikos databases for articles published within the last 5 years up till March 2023. Eligible studies were systematic reviews that stated misinformation or related terms as part of their objectives, conducted a systematic search of at least one database, and reported at least 1 definition for misinformation or related terms. We extracted definitions for the terms misinformation, disinformation, fake news, infodemic, and malinformation. Within each definition, we identified concepts and mapped them across misinformation-related terms. We included 41 eligible systematic reviews, out of which 32 (78%) reviews addressed the topic of public health emergencies (including the COVID-19 pandemic) and contained 75 definitions for misinformation and related terms. The definitions consisted of 20 for misinformation, 19 for disinformation, 10 for fake news, 24 for infodemic, and 2 for malinformation. \"False/inaccurate/incorrect\" was mentioned in 15 of 20 definitions of misinformation, 13 of 19 definitions of disinformation, 5 of 10 definitions of fake news, 6 of 24 definitions of infodemic, and 0 of 2 definitions of malinformation. Infodemic had 19 of 24 definitions addressing \"information overload\" and malinformation had 2 of 2 definitions with \"accurate\" and 1 definition \"used in the wrong context.\" Out of all the definitions, 56 (75%) were referenced from other sources. While the definitions of misinformation and related terms in the health field had inconstancies and variability, they were largely consistent. Inconstancies related to the intentionality in misinformation definitions (7 definitions mention \"unintentional,\" while 5 definitions have \"intentional\"). They also related to the content of infodemic (9 definitions mention \"valid and invalid info,\" while 6 definitions have \"false/inaccurate/incorrect\"). The inclusion of concepts such as \"intentional\" may be difficult to operationalize as it is difficult to ascertain one's intentions. This scoping review has the strength of using a systematic method for retrieving articles but does not cover all definitions in the extant literature outside the field of health. This scoping review of the health literature identified several definitions for misinformation and related terms, which showed variability and included concepts that are difficult to operationalize. Health practitioners need to exert caution before labeling a piece of information as misinformation or any other related term and only do so after ascertaining accurateness and sometimes intentionality. Additional efforts are needed to allow future consensus around clear and operational definitions.

Approximately 80% of internet users access health information online and patients with chronic illnesses especially rely on internet-based resources. YouTube ranks second among the most accessed websites worldwide and hosts an increasing number of videos with medical information. However, their quality is sometimes unscientific, misleading, or even harmful. As little is known about YouTube as a source of information on psoriasis, we aimed to investigate the quality of psoriasis-related videos and, if necessary, point out strategies for their improvement. The quality of the 100 most viewed psoriasis-related videos was assessed using the DISCERN instrument and the Global Quality Scale (GQS) by categorizing the videos into useful, misleading, and dangerous and by evaluating the reception of the videos by users. Evaluation of the videos exhibited a total of 117,221,391 views and a total duration of 10:28 hour. The majority of clips contained anecdotal personal experiences with complementary and alternative psoriasis treatments, topical treatments, and nutrition and diets being the most frequently addressed topics. While advertisements accounted for 26.0% (26/100) of the videos, evidence-based health information amounted to only 20.0% (20/100); 32.0% (32/100) of the videos were classified as useful, 52.0% (52/100) as misleading, and 11.0% (11/100) as even dangerous. The quality of the videos evaluated by DISCERN and GQS was generally low (1.87 and 1.95, respectively, on a 1 to 5 scale with 5 being the maximum). Moreover, we found that viewers rated poor-quality videos better than higher quality videos. Our in-depth study demonstrates that nearly two-thirds of the psoriasis-related videos we analyzed disseminate misleading or even dangerous content. Subjective anecdotal and unscientific content is disproportionately overrepresented and poor-quality videos are predominantly rated positively by users, while higher quality video clips receive less positive ratings. Strategies by professional dermatological organizations are urgently needed to improve the quality of information on psoriasis on YouTube and other social media.

The functional food market has experienced significant growth, leading to an uptick in clinical trials conducted by contract research organizations (CROs). Research focusing on CRO-managed trials and the communication of trial outcomes to the consumer market remains underexplored. This metaepidemiological study aims to evaluate the quality of randomized controlled trials (RCTs) facilitated by prominent CROs in Japan and to examine the quality of the representations used to convey their results to consumers. This study focused on the food trials that were registered in the University Hospital Medical Information Network Clinical Trial Registry or the International Clinical Trials Registry Platform by the top 5 CROs. Press releases of study results or advertisements of food products based on the study results were identified by conducting a Google search. The risk of bias in the RCT publications was independently assessed by 2 reviewers, who also evaluated the presence of “spin” in the abstracts and full texts. An assessment of “spin” in press releases/advertisements was undertaken. A total of 76 RCT registrations, 32 RCT publications, and 11 press releases/advertisements were included. Approximately 72% of the RCT publications exhibited a high risk of bias due to selective outcome reporting. “Spin” was present in the results of the abstract (72%), abstract conclusion (81%), full-text results (44%), and full-text conclusion (84%). “Spin” appeared in 73% of press releases/advertisements due to the selective outcome reporting. Functional food presentations in Japan frequently contained “spin.” The Japanese government should more rigorously check whether food manufacturers report outcomes selectively.

Background The propagation of tobacco-related information that is inconsistent with public health guide significantly impacts public health, particularly affecting people with less access to reliable information sources (such as those with lower education), who may also suffer disproportionate tobacco-related morbidity and mortality. This study develops a multi-dimensional analytical framework for identifying and categorizing tobacco-related information on social media. Using a dataset of tweets, the framework was constructed through qualitative analysis, which was then compared with an exploratory, AI-assisted analysis to assess the capabilities of current automated tools. Methods A collection of 3.4 million tweets related to tobacco and nicotine was refined to 842,754 after removing irrelevant and duplicate posts. LDA topic modeling identified six unique topics, from which two randomly selected samples of tweets were drawn to perform qualitative analysis and AI-assisted analysis to identify categories of tobacco information. Results The identified tobacco-related information was categorized by three dimensions (1) content, including safety and health effects, cessation, substance, and policy; (2) type of falsehood, which included fabrication and unsubstantiated claims, misrepresentations, and distortions; and (3) source, ranging from individuals and retail stores to advocacy groups and influencers. A notable finding was the prevalence of policy-related discussions of tobacco information on Twitter (X), highlighting this often-overlooked domain. The controversy over vaping has amplified pro-vaping voices on social media, with content frequently misinterpreting scientific findings, policies, and expert opinions, reflecting more nuanced and difficult to recognize falsehood in the misleading content. Conclusion This study offers a comprehensive framework for analyzing tobacco-related information on social media, emphasizing key issues in policy debates and the presence of conspiracy narratives. This framework can inform the design of interventions for less informed populations and enhance data annotation for machine learning tasks.

ABSTRACT BACKGROUND False and misleading advertising for drugs can harm consumers and the healthcare system, and previous research has demonstrated that physician-targeted drug advertisements may be misleading. However, there is a dearth of research comparing consumer-targeted drug advertising to evidence to evaluate whether misleading or false information is being presented in these ads. OBJECTIVE To compare claims in consumer-targeted television drug advertising to evidence, in order to evaluate the frequency of false or misleading television drug advertising targeted to consumers. DESIGN A content analysis of a cross-section of television advertisements for prescription and nonprescription drugs aired from 2008 through 2010. We analyzed commercial segments containing prescription and nonprescription drug advertisements randomly selected from the Vanderbilt Television News Archive, a census of national news broadcasts. MAIN MEASURES For each advertisement, the most-emphasized claim in each ad was identified based on claim iteration, mode of communication, duration and placement. This claim was then compared to evidence by trained coders, and categorized as being objectively true, potentially misleading, or false. Potentially misleading claims omitted important information, exaggerated information, made lifestyle associations, or expressed opinions. False claims were factually false or unsubstantiated. KEY RESULTS Of the most emphasized claims in prescription ( n  = 84) and nonprescription ( n  = 84) drug advertisements, 33 % were objectively true, 57 % were potentially misleading and 10 % were false. In prescription drug ads, there were more objectively true claims (43 %) and fewer false claims (2 %) than in nonprescription drug ads (23 % objectively true, 7 % false). There were similar numbers of potentially misleading claims in prescription (55 %) and nonprescription (61 %) drug ads. CONCLUSIONS Potentially misleading claims are prevalent throughout consumer-targeted prescription and nonprescription drug advertising on television. These results are in conflict with proponents who argue the social value of drug advertising is found in informing consumers about drugs.

Despite significant developments in Aboriginal and Torres Strait Islander Health information over the last 25 years, many challenges remain. There are still uncertainties about the accuracy of estimates of the summary measure of life expectancy, and methods to estimate changes in life expectancy over time are unreliable because of changing patterns of identification. Far too little use is made of the wealth of information that is available, and formal systems for systematically using that information are often vestigial to non-existent. Available information has focussed largely on traditional biomedical topics and too little on access to, expenditure on, and availability of services required to improve health outcomes, and on the underpinning issues of social and emotional wellbeing. It is of concern that statistical artefacts may have been misrepresented as indicating real progress in key health indices. Challenges and opportunities for the future include improving the accuracy of estimation of life expectancy, provision of community level data, information on the availability and effectiveness of health services, measurement of the underpinning issues of racism, culture and social and emotional wellbeing (SEWB), enhancing the interoperability of data systems, and capacity building and mechanisms for Indigenous data governance. There is little point in having information unless it is used, and formal mechanisms for making full use of information in a proper policy/planning cycle are urgently required.