Information Bias Research Articles

Does the association between physical activity during work and leisure and blood pressure differ across sex? A cross-sectional compositional data analysis in a Danish population-based cohort.

A proposed risk factor for cardiovascular disease is high occupational physical activity (OPA), which seems to increase blood pressure (BP), in contrast to leisure time physical activity. Increased BP may lead to hypertension that increase the risk of cardiovascular disease and premature death. Exposures to OPA differ across sex and also within occupational group. Thus, we aimed to investigate associations between OPA and LTPA and BP among men and women using compositional data analysis. This population-based cross-sectional study, used data from the Copenhagen Aging and Midlife Biobank. OPA and LTPA were self-reported time spent in light physical activity (LPA) during work = standing or walking work; moderate-to-vigorous physical activity (MVPA) during work = heavy manual work; LPA during leisure = light physical activity during leisure; MVPA during leisure = biking or walking as commute to work + daily amount of MVPA during leisure, and sleep. Systolic and diastolic BP (SBP, DBP; mmHg) was measured during sitting rest. We used linear regression models to investigate the association between OPA and LTPA, expressed as isometric log-ratios, and BP. The models were used to predict the BP for reallocated physical activity (PA) compositions (i.e., theoretically 'moving' time from sitting to PA within each domain). Specifically, we predicted the BP for each reallocated PA compositions and calculated the difference in BP between the reallocated compositions and the mean composition. In total, 1,334 women and 2,983 men (mean age 55.1 and 52.5 years, respectively) were included in the analyses. About 50% of the women, and 66% of the men, had hypertension. The linear regressions based on the compositional data analysis, showed no association between OPA and LTPA and SBP among women or men. Among men, less time spent sitting and more time spent in LTPA, compared to the mean composition, was associated with a lower DBP (e.g., 60min less sitting and 60min more LTPA: -0.25, 95% CI: -0.05, -0.45 mmHg). No association between OPA and LTPA and BP was observed across sexes, except between LTPA and DBP among men. This could be due to information bias and lack of precision in self-reported time use data of PA. None.

Sources of bias in studies reporting birth prevalence of congenital anomalies: a scoping review and reporting checklist

Sources of bias in studies reporting birth prevalence of congenital anomalies: a scoping review and reporting checklist

Modern Sources of Controls in Case-Control Studies

Modern Sources of Controls in Case-Control Studies

Breaking up Big Tech: scissor-line suggestions for smart cuts

Abstract Calls for the breakup of Big Tech, especially where behavioural remedies are cumbersome to enforce effectively, are becoming louder on both sides of the Atlantic. If there are going to be structural remedies against network companies, we propose they better be ‘smart cuts’: targeted interventions to insulate core platform functions with minimal damage to positive network effects. One major concern with dominant platforms is information bias: intentional distortion of information presented to a user to steer them away from the best matches to their query and towards content that benefits the platform instead. It can be addressed by separating a hybrid platform’s recommender system from its peripheral services with commercial tentacles in the real economy. This realigns incentives to provide users with unbiased information and invites competition on and between intermediaries. We offer scissor-line suggestions for the gatekeepers Alphabet, Amazon, and Apple. Smart cuts have a conceptual basis: they go beyond unwinding recent acquisitions and open up ecosystem structures to competition, allow for targeted flanking access regulation, and alleviate competition law enforcement.

The predictive effect of ASD on PTSD and the factors influencing ASD and PTSD

ObjectiveTo investigate the prevalence and influencing factors of acute stress disorder (ASD) and post-traumatic stress disorder (PTSD) in trauma patients, and to explore the predictive effect of ASD on PTSD. MethodsA prospective study was conducted on patients hospitalized due to injuries. The first survey used the ASD scale to assess the occurrence of ASD. In one month and three months of follow-up after injury, patients were assessed for the occurrence of PTSD by using the PTSD checklist-civilian version. ResultsThe prevalence rates of ASD, one-month PTSD, and three-month PTSD in trauma inpatients were 20.7%, 19.5%, and 17.6%, respectively. ASD is a strong predictor of PTSD, and combining it with severe injury and critical illness can improve the sensitivity and positive predictive ability of predicting the occurrence of PTSD (AUCMax: 0.827). The important predictive factor for the diagnosis of PTSD is the high alert symptom group of ASD. Moreover, the analysis results showed that the season of trauma happened, comatose state, fear state, psychological burden, and pain intensity were the influencing factors for ASD (P<0.05), while critical illness during hospitalization, psychological burden, and pain intensity were the influencing factors for PTSD (P<0.05). LimitationsSome patients with minor and extremely serious injuries were overlooked or missed, resulting in selection bias and information bias that could not be completely avoided. ConclusionBoth trauma conditions and clinical features may affect the occurrence of ASD and PTSD in trauma patients. If ASD in trauma patients is identified early and targeted interventions, it may reduce the occurrence and development of PTSD.

Fact-checking in the age of AI: Reducing biases with non-human information sources

This study examines the obstacles to the effectiveness of fact-checking, focusing primarily on the pervasive impact of entrenched biases. Fact-checking efforts often face resistance when linked to mistrusted sources, leading to cognitive dissonance and the rejection of messages in favor of pre-existing beliefs, a phenomenon known as motivated reasoning. This resistance hinders organizations’ ability to correct misconceptions surrounding social issues and entities. The research delves into whether non-human entities such as AI can facilitate less biased information processing due to their perceived impartiality. Applying a moderated mediation model in experimental settings, we found that labeling a source as artificial intelligence is pivotal in evaluating fact-checking. AI labels moderate the impact of partisan biases on the persuasive outcomes of fact-checks, such as message credibility and acceptance, compared to the human source. This study offers valuable insights for enhancing the effectiveness of fact-checking in the context of cognitive and psychological biases by highlighting the critical influence of information sources in reducing polarization in public perceptions of scientific issues.

Abstract A027: Systematic review of real-world adherence to blood-based laboratory tests for average-risk disease screening: potential implications for emerging colorectal cancer screening modalities

Abstract Introduction: Blood-based tests represent a compelling option for early detection and management of cancers and other chronic diseases. While blood-based tests may offer potential to increase patient engagement, the assumption of greater adherence in clinical practice requires further evaluation. This systematic review aimed to evaluate the real-world adherence to established blood-based tests for commonly recommended screening indications, to better inform expectations for proposed application to average-risk colorectal cancer (CRC) screening. Methods: A systematic search of PubMed, Embase, and citations was performed to identify literature published from 2010 to 2023. Studies examined adherence to routine blood-based laboratory tests with a Grade A or B recommendation per the United States Preventive Services Task Force (USPSTF) for diabetes, dyslipidemia, human immunodeficiency virus (HIV), and hepatitis C virus (HCV) in an average-risk patient population were included. In addition, prostate-specific antigen (PSA) testing adherence for prostate cancer screening (Grade-C USPSTF recommendation), the only commonly accessible blood-based single-cancer screening test for average-risk adults, was also included. Studies without venipuncture-derived blood were excluded. Results: Of 53,067 articles screened, 69 were included in the analysis. Overall, adherence rates of blood-based tests were highly variable, with median values (interquartile range, IQR) of 66.3% (11.8%), 67.8% (18.0%), 34% (28.7%), and 36.8% (30%) for diabetes, dyslipidemia, HCV, and HIV, respectively. PSA testing adherence was 37.2% (18.4%). Beyond disease type, adherence varied by study design, data source, patient characteristics, clinical setting, and study country. Information and selection bias were the most common risks of bias. Conclusions: Based on data from this comprehensive systematic review, real-world adherence to currently recommended, blood-based screening strategies is suboptimal. Future research is needed to identify the extent to which these data are generalizable to blood-based CRC screening, and to determine if/how this candidate strategy would impact clinical, economic, and health equity outcomes. Citation Format: Quang A. Le, Takako Kiener, Heather A. Johnson, Rukaiyya S. Khan, Joyce Kong, Kevin H. Li, Gina Thompson, A. Burak Ozbay, Mallik Greene, Paul J. Limburg, A. Mark Fendrick, John B. Kisiel, Derek W. Ebner. Systematic review of real-world adherence to blood-based laboratory tests for average-risk disease screening: potential implications for emerging colorectal cancer screening modalities [abstract]. In: Proceedings of the AACR Special Conference: Liquid Biopsy: From Discovery to Clinical Implementation; 2024 Nov 13-16; San Diego, CA. Philadelphia (PA): AACR; Clin Cancer Res 2024;30(21_Suppl):Abstract nr A027.

The Potential of Artificial Intelligence Tools for Reducing Uncertainty in Medicine and Directions for Medical Education.

In the field of medicine, uncertainty is inherent. Physicians are asked to make decisions on a daily basis without complete certainty, whether it is in understanding the patient's problem, performing the physical examination, interpreting the findings of diagnostic tests, or proposing a management plan. The reasons for this uncertainty are widespread, including the lack of knowledge about the patient, individual physician limitations, and the limited predictive power of objective diagnostic tools. This uncertainty poses significant problems in providing competent patient care. Research efforts and teaching are attempts to reduce uncertainty that have now become inherent to medicine. Despite this, uncertainty is rampant. Artificial intelligence (AI) tools, which are being rapidly developed and integrated into practice, may change the way we navigate uncertainty. In their strongest forms, AI tools may have the ability to improve data collection on diseases, patient beliefs, values, and preferences, thereby allowing more time for physician-patient communication. By using methods not previously considered, these tools hold the potential to reduce the uncertainty in medicine, such as those arising due to the lack of clinical information and provider skill and bias. Despite this possibility, there has been considerable resistance to the implementation of AI tools in medical practice. In this viewpoint article, we discuss the impact of AI on medical uncertainty and discuss practical approaches to teaching the use of AI tools in medical schools and residency training programs, including AI ethics, real-world skills, and technological aptitude.

Uncovering coordinated cross-platform information operations: Threatening the integrity of the 2024 U.S. presidential election

Information operations (IOs) pose a significant threat to the integrity of democratic processes, with the potential to influence election-related online discourse. In anticipation of the 2024 U.S. presidential election, we present a study aimed at uncovering the digital traces of coordinated IOs on X (formerly Twitter). Using our machine learning framework for detecting online coordination, we analyze a dataset comprising election-related conversations on X from May to July 2024. This reveals a network of coordinated inauthentic actors, displaying notable similarities in their link-sharing behaviors. Our analysis shows concerted efforts by these accounts to disseminate misleading, redundant, and biased information across the Web through a coordinated cross-platform information operation: The links shared by this network frequently direct users to other social media platforms or mock news sites featuring low-quality political content and, in turn, promoting the same X and YouTube accounts. Members of this network also shared deceptive images generated by AI, accompanied by language attacking political figures and symbolic imagery intended to convey power and dominance. While X has suspended or restricted a subset of these accounts, 75 percent ofthe coordinated network remains active, garnering substantial traction over time: The suspicious Web sites promoted by this coordinated network are shared thousands of times per day by the X user base, further amplifying their reach and potential impact. Our findings underscore the critical role of developing computational models to scale up the detection of threats on large social media platforms, and emphasize the broader implications of these techniques to detect IOs across the wider Web.

Agile, on-demand wastewater surveillance of virus infections to support pandemic and outbreak response in Rotterdam-Rijnmond, the Netherlands, 2020 to 2022.

BackgroundWastewater surveillance may support early and comprehensive detection of infectious diseases' community transmission, particularly in settings where other health surveillance systems provide biased or limited information. Amid the SARS-CoV-2 pandemic, deploying passive samplers to monitor targeted populations gained importance. Evaluation of the added public health value of this approach in the field can support its broader adoption.AimWe aimed to assess the feasibility and utility of on-demand wastewater surveillance, employing passive samplers, for SARS-CoV-2 and monkeypox virus (MPXV) in small/targeted populations, also considering ethical aspects.MethodsPilot case studies in the Rotterdam-Rijnmond region were used for a systematic assessment of the feasibility and utility of wastewater monitoring of SARS-CoV-2 (variants) and MPXV using passive sampling. Each case study was instigated by actual questions from the Public Health Service about disease transmission.ResultsCase study results demonstrated the feasibility and utility of on-demand wastewater surveillance with successful identification of a local peak in SARS-CoV-2 transmission, early detection of wider Omicron variant transmission after the first case was reported, as well as indication of no emerging local MPXV transmission. Ethical considerations led to the abandonment of one case study involving a displaced population.ConclusionsThe study confirms the feasibility and utility of passive sampling for real-time infectious disease surveillance, at desired spatiotemporal resolution. Ethical concerns and operational challenges were identified, highlighting the need for early stakeholder engagement and ethical guideline adherence. The method could be used to study under-surveyed populations and be extended beyond SARS-CoV-2 and MPXV to other pathogens.

Public Perceptions of Male Hormone Levels and Prostate Health: A Text Mining Analysis of South Korean Media and Online Communities.

The purpose of this study is to identify social perceptions and public concerns related to male hormone levels and prostate cancer. To do so, we analyzed news and community data from Naver, Daum, and Google, the three most active social media platforms in South Korea, to explore how male hormone levels and benign prostatic hyperplasia (BPH) are covered in the media and discussed within online communities. We analyzed three years of data from community discussions and news articles from July 2021 to June 2024. Text mining techniques were used to collect data with the keyword 'male hormone levels' to understand the portrayal of these issues in news and online communities as well as the public's social perceptions and concerns. The analysis showed that news coverage of 'male hormone levels' initially drew public attention by highlighting information related to celebrities and events while discussion on 'prostate enlargement' largely focused on medical treatments and information. The online community served as a platform where news-triggered issues were rediscussed through personal experiences and opinions, thus influencing public perceptions. However, the community discussions leaned more towards sexual dysfunction rather than prostate enlargement. These findings suggest that while news and online communities shape public awareness in different way, they both contribute to shaping social understanding of male hormone level and BPH. This study highlights the significant roles that news and online communities play in shaping social perceptions of male hormone levels and BPH. However, sensationalism in media coverage may lead to biased information and reinforce false stereotypes. Future studies can be done to incorporate data from diverse cultural contexts and extend the data collection period to further explore shifts in public perceptions and the evolution of social discussions surrounding male hormone levels and BPH.

Reducing Information and Selection Bias in EHR-Linked Biobanks via Genetics-Informed Multiple Imputation and Sample Weighting.

Electronic health records (EHRs) are valuable for public health and clinical research but are prone to many sources of bias, including missing data and non-probability selection. Missing data in EHRs is complex due to potential non-recording, fragmentation, or clinically informative absences. This study explores whether polygenic risk score (PRS)-informed multiple imputation for missing traits, combined with sample weighting, can mitigate missing data and selection biases in estimating disease-exposure associations. Simulations were conducted for missing completely at random (MCAR), missing at random (MAR), and missing not at random (MNAR) conditions under different sampling mechanisms. PRS-informed multiple imputation showed generally lower bias, particularly when combined with sample weighting. For example, in biased samples of 10,000 with exposure and outcome MAR data, PRS-informed imputation had lower percent bias (3.8%) and better coverage rate (0.883) compared to PRS-uninformed (4.5%; 0.877) and complete case analyses (10.3%; 0.784) in covariate-adjusted, weighted, multiple imputation scenarios. In a case study using Michigan Genomics Initiative (n=50,026) data, PRS-informed imputation aligned more closely with a sample-weighted All of Us-derived benchmark than analyses ignoring missing data and selection bias. Researchers should consider leveraging genetic data and sample weighting to address biases from missing data and non-probability sampling in biobanks.

School ties and evaluation outcomes: Evidence from the Korean Basketball League

AbstractThis paper investigates whether evaluation outcomes are influenced by private information and subjective biases when appraisers and appraisees are socially connected. In the Korean Basketball League, school ties between referees and players are commonplace because a large proportion of referees were previously basketball players. Using data from six basketball seasons, we analyze the degree to which referees' decisions are affected when players and referees attended the same school. Our results suggest that players who play under referees with school ties are called for fewer fouls. We also find that the results are driven mainly by high school ties.

Association between Thai language proficiency and adherence to COVID-19 protective behaviors (CPB) among Myanmar migrant workers in Southern Thailand.

The association between host country language proficiency and disease prevention among migrants is underexplored. The objective of this study is to assess the extent to which self-reported command of the Thai language is associated with adherence to COVID-19 protective behaviors (CPB) among Myanmar migrant workers in Thailand. We distributed a self-administered structured questionnaire in Burmese language to 1,050 Myanmar migrant workers in Southern Thailand from September 2022 to January 2023. The questionnaire included background characteristics, self-reported Thai language proficiency based on the Common European Framework Reference (CEFR), and self-reported CPB adherence at residence and workplace. We analyzed data using descriptive statistics and multivariate linear regression analysis. Although slightly less than half of the participants reported CEFR A1 level or higher in Thai speaking and listening skills, less than 10 percent did so for reading and writing skills. Workplace COVID-19 preventive adherence scores were initially found to be significantly associated with A1 level or higher speaking and listening skills. However, after adjusting for confounders, these associations were not statistically significant (Speaking skill's Adjusted Beta = 0.713, 95% CI = -0.011, 1.437; Listening skill's Adjusted Beta = -0.367, 95% CI = -1.087, 0.353). No significant associations were found between any language skill domain and residence COVID-19 preventive adherence scores for both unadjusted and adjusted analysis. The study findings may have implications for relevant stakeholders in migrant services, migrant health, and infectious disease control. However, information biases, language barriers, and lack of generalizability should be considered as caveats in the interpretation of the study findings.

Biased, Not Balanced Broadcaster! Deconstructing Bias Accusations Toward Public Service Media

Across Europe, Public Service Media (PSM) are increasingly subject to bias accusations in public debates. Excluding academic attention for actual bias in news content, research into the nature of bias accusations is limited. This article studies bias accusations against Dutch PSM in online discourse. Through a qualitative content analysis of tweets and blogs (2017–2022), our analysis develops a novel typology of bias accusations—positional, information, and framing bias—going beyond academic definitions. Bias accusations are diverse in nature, reflect strong perceived intentionality, and refer to heterogeneous social, political groups, yet are easily merged into singular political ideological divides.

Emergence of information aggregation to rational expectations equilibria in markets populated by biased heuristic traders

Information aggregation is a key economic function of markets. We report results of a computational experiment with markets populated by simple algorithmic traders who follow two heuristics usually thought of as leading to biased information processing in behavioral economics literature (anchor-and-adjust, and representativeness). Outcomes of these markets either tend to cluster around (or fail to do so) rational expectations equilibria under specific conditions, consistent with markets populated by profit-motivated human traders. Algorithmic trader convergence is slower and noisier than that of human traders. Our results illustrate the emergence of rational expectations equilibria through complex interactions among actions of biased heuristic traders with limited information processing capabilities.

Accuracy, Quality, and Misinformation of YouTube Abortion Procedural Videos: Cross-Sectional Study.

The internet is often the first source patients turn to for medical information. YouTube is a commonly used internet-based resource for patients seeking to learn about medical procedures, including their risks, benefits, and safety profile. Abortion is a common yet polarizing medical procedure. People interested in obtaining an abortion are likely to use the internet to learn more about abortion procedures and may encounter misinformed and biased information. This is troubling as information found on the internet can significantly alter perceptions and understanding of these procedures. There is no current research that evaluates the accuracy, quality, and misinformation of instructional abortion videos available to patients. The purpose of this study was to assess if any given video can deliver accurate and quality information about this topic in an unbiased manner and to assess the level of factually incorrect, distorted, or medically irrelevant information in any given video. Procedural methods of abortion were queried on YouTube on August 22, 2022. The videos were screened with strict exclusion criteria. Videos were categorized into "video slants" based on the language and attitudes expressed in each video. Video accuracy was calculated using the Surgical Curriculum in Obstetrics and Gynecology (SCOG) checklist for each corresponding procedure. Video quality was calculated using the Laparoscopic Surgery Video Educational Guidelines (LAP-VEGaS) criteria. The level of misinformation was assessed with the evidence-based Anti-Choice Rubric, which scores the amount of factually incorrect, distorted, or medically irrelevant information in each video. A total of 32 videos were analyzed and categorized into 3 "video slant" groups: neutral (n=23, 72%), antichoice (n=4, 12%), and prochoice (n=5, 16%). Using the SCOG checklist, neutral videos had the highest median accuracy (45.9%), followed by antichoice videos (24.6%) and prochoice videos (18.5%). None of the videos met the LAP-VEGaS quality control criteria, (score>11, indicating adequate quality). Neutral videos had a median score of 8.8 out of 18, with antichoice videos scoring 10.75 and prochoice videos scoring 6.2. Using the Anti-Choice Rubric, neutral videos mentioned only 1 factually incorrect piece of information. Antichoice videos mentioned 12 factually incorrect pieces of information, 8 distortions, and 3 medically irrelevant pieces of information. Prochoice videos did not mention any of the 3 themes. Using the SCOG checklist, the accuracy of instructional videos were inconsistent across the 3 identified "video slants." Using LAP-VEGaS criteria, the quality of educational videos were also inconsistent across the 3 "video slants." Prochoice videos had the lowest level of misinformation, with no mentions of any of the 3 themes. Antichoice videos had the highest levels of misinformation, with mentions in all 3 themes. Health care professionals should consider this when counseling patients who may watch YouTube videos for information regarding abortion procedures.

Learnable Transform-Assisted Tensor Decomposition for Spatio-Irregular Multidimensional Data Recovery

Tensor decompositions have been successfully applied to multidimensional data recovery. However, classical tensor decompositions are not suitable for emerging spatio-irregular multidimensional data (i.e., spatio-irregular tensor), whose spatial domain is non-rectangular, e.g., spatial transcriptomics data from bioinformatics and semantic units from computer vision. By using preprocessing (e.g., zero-padding or element-wise 0-1 weighting), the spatio-irregular tensor can be converted to a spatio-regular tensor and then classical tensor decompositions can be applied, but this strategy inevitably introduces bias information, leading to artifacts. How to design a tensor-based method suitable for emerging spatio-irregular tensors is an imperative challenge. To address this challenge, we propose a learnable transform-assisted tensor singular value decomposition (LTA-TSVD) for spatio-irregular tensor recovery, which allows us to leverage the intrinsic structure behind the spatio-irregular tensor. Specifically, we design a learnable transform to project the original spatio-irregular tensor into its latent spatio-regular tensor, and then the latent low-rank structure is captured by classical TSVD on the resulting regular tensor. Empowered by LTA-TSVD, we develop spatio-irregular low-rank tensor completion (SIR-LRTC) and spatio-irregular tensor robust principal component analysis (SIR-TRPCA) models for the spatio-irregular tensor imputation and denoising respectively, and we design corresponding solving algorithms with theoretical convergence. Extensive experiments including the spatial transcriptomics data imputation and hyperspectral image denoising show SIR-LRTC and SIR-TRPCA are superior performance to competing approaches and benefit downstream applications.

Evaluation of Children with Familial Mediterranean Fever in South-Eastern Region of Türkiye

Background: Familial Mediterranean fever (FMF) is a hereditary autoinflammatory disorder commonly found in populations from the Mediterranean region, including Turks, Armenians, Arabs, and Jews. It is characterized by recurrent episodes of fever and inflammation, primarily affecting the peritoneum, pleura, and joints. Familial Mediterranean fever is caused by mutations in the MEFV gene, which encodes pyrin, a protein that regulates the inflammatory response. The most common mutations associated with FMF are M694V, M680I, and V726A. These mutations lead to the overproduction of interleukin-1β (IL-1β), triggering the inflammatory episodes that are the hallmark of the disease. Objectives: This study aims to explore the clinical implications of MEFV gene mutations in a cohort of patients diagnosed with FMF and treated with colchicine. The primary objective is to determine the frequency of various MEFV mutations in the patient cohort and examine their association with clinical presentations, such as symptom severity and response to treatment. Methods: A retrospective cohort of 302 patients diagnosed with FMF and undergoing colchicine treatment was examined. Demographic data, clinical symptoms, and genetic mutations were systematically collected and analyzed. Mutation analysis focused on identifying the most common MEFV variants and their potential influence on clinical outcomes. Consanguinity and family history were also considered for their role in FMF genetic predisposition. Results: The study included 157 females and 145 males, with a mean age of 142.60 ± 48.04 months at the time of analysis. The average age of diagnosis was 94 ± 42 months, with no significant differences between genders. Abdominal pain was the most common symptom (90.3%), followed by fever (69.0%) and arthralgia (42.3%). Consanguinity was reported in 35.1% of cases, mainly between first-degree cousins, and 24.5% had a family history of FMF. Genetic analysis revealed that the M694V mutation was the most common homozygous mutation (71.4%), followed by the E148Q mutation as the most frequent heterozygous mutation (37.5%). The compound heterozygous mutation M694V/E148Q was found in 20% of cases. A statistically significant association was found between arthritis and homozygous mutations (P = 0.001) as well as with the M694V mutation (P = 0.004). Homozygous M694V mutations were detected in 16.9% of patients. The findings underscore a strong correlation between the M694V mutation and the development of FMF. Despite the efficacy of colchicine in managing symptoms and preventing complications such as amyloidosis, some patients exhibited treatment resistance, particularly those with arthritis. Conclusions: This study highlights the significant role of MEFV gene mutations, particularly M694V, in the pathogenesis of FMF. It underscores the importance of early diagnosis and the potential for personalized treatment based on genetic findings. While the retrospective design of the study presents limitations, such as selection and information bias, the results provide valuable insights into the clinical and genetic dimensions of FMF. Future studies should build on these findings by using prospective designs and investigating the interaction between genetic and environmental factors in FMF. Expanding the scope of genetic analysis could also lead to the discovery of new mutations, enhancing our understanding of this complex disorder.

A robust visual question answering approach to reduce multimodal bias

Currently, many visual question answering models have bias problems. Specifically, when the question-answer relationship in the training data shows a more obvious mapping relationship, the model shows poor generalization ability. For such biased predictions, existing research work mainly considers language bias, while ignoring the bias information introduced by images. In order to enhance the robustness of visual question answering models, a bias reduction method is proposed, and on this basis, the influence of language and visual information on bias is explored. Furthermore, two bias learning branches are constructed to capture language bias and the bias caused by language and images respectively, and the bias reduction method is used to obtain more robust prediction results. Finally, according to the difference in prediction probability between the standard visual question answering and bias branches, the samples are dynamically weighted, so that the model can dynamically adjust the learning degree for samples with different bias levels. Experiments on datasets such as VQA-CP v2.0 prove the effectiveness of the proposed method and alleviate the influence of bias on the model.