Search

Background: A sex-based evaluation of prognosis in heart failure (HF) is lacking. Methods and results: We analyzed the Metabolic Exercise test data combined with Cardiac and Kidney Indexes (MECKI) score registry, which includes HF with reduced ejection fraction (HFrEF) patients. A cross-validation procedure was performed to estimate weights separately for men and women of all MECKI score parameters: left ventricular ejection fraction (LVEF), hemoglobin, kidney function assessed by Modification of Diet in Renal Disease, blood sodium level, ventilation vs. carbon dioxide production slope, and peak oxygen consumption (peakVO2). The primary outcomes were the composite of all-cause mortality, urgent heart transplant, and implant of a left ventricle assist device. The difference in predictive ability between the native and sex recalibrated MECKI (S-MECKI) was calculated using a receiver operating characteristic (ROC) curve at 2 years and a calibration plot. We retrospectively analyzed 7,900 HFrEF patients included in the MECKI score registry (mean age 61 ± 13 years, 6,456 men/1,444 women, mean LVEF 33% ± 10%, mean peakVO2 56.2% ± 17.6% of predicted) with a median follow-up of 4.05 years (range 1.72–7.47). Our results revealed an unadjusted risk of events that was doubled in men compared to women (9.7 vs. 4.1) and a significant difference in weight between the sexes of most of the parameters included in the MECKI score. S-MECKI showed improved risk classification and accuracy (area under the ROC curve: 0.7893 vs. 0.7799, p = 0.02) due to prognostication improvement in the high-risk settings in both sexes (MECKI score >10 in men and >5 in women). Conclusions: S-MECKI, i.e., the recalibrated MECKI according to sex-specific differences, constitutes a further step in the prognostic assessment of patients with severe HFrEF.

Around 13% of fossil fuels globally are used for non-combustion purposes. Fossil fuel processing plants, such as petroleum refineries, exhibit interdependent material and energy system dynamics, making the transition away from fossil fuels in energy systems more challenging without addressing the non-energy outputs. This study explores the future role of fossil fuels for non-energy purposes in climate-stringent scenarios with restrictions on alternative feedstock availability, focusing on the primary chemicals sector. Using a global integrated assessment model with detailed refining and primary chemicals sectors, findings across various scenarios reveal that up to 62% of total feedstock use in the chemical sector could be provided by alternative sources by 2050. This would require significant scale-up in biomass utilisation and carbon capture technologies. Annual CO2 emissions from the chemical sector could be reduced to as low as -1Gt CO2 by the same year if carbon storage in non-recycled and non-incinerated bioplastics is accounted for.

Despite faster-than-expected progress in clean energy technology deployment, global annual CO2 emissions have increased from 2020 to 2023. The feasibility of limiting warming to 1.5 °C is therefore questioned. Here we present a model intercomparison study that accounts for emissions trends until 2023 and compares cost-effective scenarios to alternative scenarios with institutional, geophysical and technological feasibility constraints and enablers informed by previous literature. Our results show that the most ambitious mitigation trajectories with updated climate information still manage to limit peak warming to below 1.6 °C (‘low overshoot’) with around 50% likelihood. However, feasibility constraints, especially in the institutional dimension, decrease this maximum likelihood considerably to 5–45%. Accelerated energy demand transformation can reduce costs for staying below 2 °C but have only a limited impact on further increasing the likelihood of limiting warming to 1.6 °C. Our study helps to establish a new benchmark of mitigation scenarios that goes beyond the dominant cost-effective scenario design.

Background and objective: Innovations have improved outcomes in advanced prostate cancer (PC). Nonetheless, we continue to lack high-level evidence on a variety of topics that greatly impact daily practice. The 2024 Advanced Prostate Cancer Consensus Conference (APCCC) surveyed experts on key questions in clinical management in order to supplement evidence-based guidelines. Here we present voting results for questions from APCCC 2024. Methods: Before the conference, a panel of 120 international PC experts used a modified Delphi process to develop 183 multiple-choice consensus questions on eight different topics. Before the conference, these questions were administered via a web-based survey to the voting panel members (“panellists”). Key findings and limitations: Consensus was a priori defined as ≥75% agreement, with strong consensus defined as ≥90% agreement. The voting results show varying degrees of consensus, as discussed in this article and detailed in the Supplementary material. These findings do not include a formal literature review or meta-analysis. Conclusions and clinical implications: The voting results can help physicians and patients navigate controversial areas of clinical management for which high-level evidence is scant or conflicting. The findings can also help funders and policymakers in prioritising areas for future research. Diagnostic and treatment decisions should always be individualised on the basis of patient and cancer characteristics, and should incorporate current and emerging clinical evidence, guidelines, and logistic and economic factors. Enrolment in clinical trials is always strongly encouraged. Importantly, APCCC 2024 once again identified important gaps (areas of nonconsensus) that merit evaluation in specifically designed trials.

Safety and maintaining high performance are key considerations during the operation of lithium-ion batteries. Battery degradation, in particular lithium plating and loss of active material, is often accelerated by fast charging. This study explores a strategy for the design of fast charging protocols that takes into account the influence of the variability between battery cells on factors that can impact degradation. We employ a non-intrusive polynomial chaos expansion to identify the key parameters for each degradation condition. We explore the reduction of battery degradation by adjusting constraints such as the maximum C-rate and voltage. Tight control of the key adjustable parameters contributes significantly to reducing the confidence interval of the degradation factors, allowing reduced charging time with minimal degradation. The application of our approach to two state-dependent fast charging protocols for a LiC6/LiCoO2 battery indicates the value in explicitly accounting for uncertainties when designing charging protocols that minimize degradation.

Despite faster-than-expected progress in clean energy technology deployment, global annual CO2 emissions have increased from 2020 to 2023. The feasibility of limiting warming to 1.5 °C is therefore questioned. Here we present a model intercomparison study that accounts for emissions trends until 2023 and compares cost-effective scenarios to alternative scenarios with institutional, geophysical and technological feasibility constraints and enablers informed by previous literature. Our results show that the most ambitious mitigation trajectories with updated climate information still manage to limit peak warming to below 1.6 °C (‘low overshoot’) with around 50% likelihood. However, feasibility constraints, especially in the institutional dimension, decrease this maximum likelihood considerably to 5–45%. Accelerated energy demand transformation can reduce costs for staying below 2 °C but have only a limited impact on further increasing the likelihood of limiting warming to 1.6 °C. Our study helps to establish a new benchmark of mitigation scenarios that goes beyond the dominant cost-effective scenario design.

In this article we report evidence from a series of semi-structured interviews with a broad sample of people living in Denmark (n = 21), about their perspectives on the future during the first months of the global Covid-19 pandemic. The thematic and discursive analyses, based on an abductive ontology, illustrate imaginings of the future along two vectors: individual to collective and descriptive to moral. On a descriptive and individual level, people imagined getting through the pandemic on a myopic day-by-day basis; on a descriptive and collective level, people imagined changes to work and socializing. Their future was bound and curtailed by their immediate present. On a moral and individual level, respondents were less detailed in their reports, but some vowed to change their behaviors. On a moral and collective level, respondents reported what the world should be like and discussed changes to environmental behaviors such as traveling, commuting, and work. The model suggests the domain of individual moral imaginings is the most difficult domain for people to imagine beyond the practicalities of their everyday lives. The implications of this model for comprehending imaginations of the future are discussed.

Social science research has produced evidence of an “integration paradox”: immigrants and their descendants who appear to have greater access to mainstream middle-class society, as indicated by their education or labor market success, often report experiencing more discrimination. Several explanations have been put forth to account for this counterintuitive pattern, supported by a wealth of empirical evidence. To assess the reliable insights generated by this line of research, this article presents a review and meta-analytic analysis of the integration paradox. The article reviews the major theoretical arguments advanced in literature, and derives crucial hypotheses concerning the phenomenon's overall nature, underlying mechanisms, and two key moderating conditions. Specifically, it discusses how factors like better education, which appear to signal improved integration into mainstream society, may also lead to greater exposure to mainstream members, heightened familiarity with exclusionary public discourse, or a heightened risk of downward social mobility. By increasing cognitive susceptibility to framing experiences in terms of discrimination or simply the opportunities to encounter discrimination, these mechanisms are expected to increase reports of discrimination. To test these propositions, the study summarizes and evaluates 280 statistical estimates published in 42 studies, utilizing recent developments in multilevel meta-analysis. The findings support the credibility of the integration paradox, particularly for immigrant minorities who are easily distinguishable from the mainstream and for those who reside in countries with a more open socio-political context of reception. However, the results also suggest that additional research is needed to fully comprehend the underlying mechanisms of the paradox.

Exercise preserves neuromuscular function in ageing through unknown mechanisms. Skeletal muscle fibroblasts (FIB) and stem cells (MuSC) are abundant in skeletal muscle and reside close to neuromuscular junctions, but their relative roles in motor neuron maintenance remain undescribed. Using direct co-cultures of embryonic rat motor neurons with either human MuSC or FIB, RNA sequencing revealed profound differential regulation of the motor neuron transcriptome, with FIB generally favoring neuron growth and cell migration and MuSC favoring production of ribosomes and translational machinery. Conditioned medium from FIB was superior to MuSC in preserving motor neurons and increasing their maturity. Lastly, we established the importance of donor age and exercise status and found an age-related distortion of motor neuron and muscle cell interaction that was fully mitigated by lifelong physical activity. In conclusion, we show that human muscle FIB and MuSC synergistically stimulate the growth and viability of motor neurons, which is further amplified by regular exercise.Competing Interest StatementThe authors have declared no competing interest.

Environmental pollution is escalating due to rapid global development that often prioritizes human needs over planetary health. Despite global efforts to mitigate legacy pollutants, the continuous introduction of new substances remains a major threat to both people and the planet. In response, global initiatives are focusing on risk assessment and regulation of emerging contaminants, as demonstrated by the ongoing efforts to establish the UN's Intergovernmental Science-Policy Panel on Chemicals, Waste, and Pollution Prevention. This review identifies the sources and impacts of emerging contaminants on planetary health, emphasizing the importance of adopting a One Health approach. Strategies for monitoring and addressing these pollutants are discussed, underscoring the need for robust and socially equitable environmental policies at both regional and international levels. Urgent actions are needed to transition toward sustainable pollution management practices to safeguard our planet for future generations.

The decarbonization of shipping has become an important policy goal. While integrated assessment models (IAMs) are often used to explore climate mitigation strategies, they typically provide little information on international shipping, which accounts for emissions of around 0.7 GtCO2 yr−1. Here we perform a multi-IAM analysis of international shipping and show the potential for decreasing annual emissions in the next decades, with reductions of up to 86% by 2050. This is primarily achieved through the deployment of low-carbon fuels. Models that represent several potential low-carbon alternatives tend to show a deeper decarbonization of international shipping, with drop-in biofuels, renewable alcohols and green ammonia standing out as the main substitutes for conventional maritime fuels. While our results align with the 2018 emission reduction goal of the International Maritime Organization, their compatibility with the agency’s revised target is still subject to a more definitive interpretation.

Introduction: Cardiovascular imaging results offer valuable information that can guide health decisions, but their impact on medication use and adherence is unclear. This systematic review and meta-analysis aimed to determine the downstream impact of cardiovascular imaging results on medication use and adherence. Methods: Searches were conducted across databases, including MEDLINE, PsychINFO, EMBASE, and relevant references up to 2024. Data were extracted from studies comparing outcomes for individuals with diseased versus normal arteries and trials comparing outcomes for individuals who were provided imaging results versus those with no access to imaging results and analysed in 2023 and 2024. Pooled odds ratios (ORs) for outcomes were calculated. Results: The analysis included 29 studies with 24 contributing data points. Initiation (OR:2.77;95% CI:1.82–4.20) and continuation (OR:2.06;95% CI:1.28–3.30) of lipid-lowering medications (LLMs), antihypertensives (OR:2.02;95% CI:1.76–2.33), and antiplatelets (OR:2.47;95% CI:1.68–3.64) were significantly higher in individuals with diseased arteries. The proportion of individuals on LLM increased by 2.7-fold in those with diseased arteries and 1.5-fold in those with normal arteries post-screening. The proportion on LLM increased by 4.2 times in the imaging group and 2.2 times in the "no imaging group" post-screening. There was a significant increase in LLM initiation (OR:2.37;95% CI: 1.17–4.79) in the imaging group, but medication continuation did not significantly differ between the imaging and "no imaging group". Discussion: Cardiovascular imaging results can prompt initiation of medications, particularly lipid-lowering medications, reflecting a proactive response to identified risk factors. However, evidence regarding medication continuation is mixed, and further research is required.

Background: While physical activity (PA) is important during youth, it is unclear if children and adolescents with developmental dysplasia of the hip (DDH), Legg-Calv & eacute;-Perthes disease (LCPD), or slipped capital femoral epiphysis (SCFE) are expected or encouraged to return to PA once they have healed. This study examines the orthopaedic surgeons' role in advising pediatric hip patients on the PA they should engage in, assessing their practice, opinions, and consensus when making recommendations. Methods: Orthopaedic surgeons were invited from member lists of 4 hip study groups. The survey included demographics, opinions regarding PA, and 10 case scenarios that queried respondents on the duration and intensity of PA as well as the restrictions on activity type that they would recommend for DDH, LCPD, or SCFE patients. Consensus was evaluated on a scale ranging from 0 to 1, with a value of 0 indicating no agreement among respondents and a value of 1 indicating complete agreement. Results: A total of 51 orthopaedic surgeons responded. While 94% agreed that it is important for school-aged hip patients to return to PA after they have healed, 53% believed that PA may compromise the hip and contribute to the development of osteoarthritis. Average standardized consensus was 0.92 for suggesting the patient engage in some amount of PA, 0.44 for suggesting the recommended daily minimum of 60 minutes of moderate-to-vigorous physical activity (MVPA), and 0.33 for suggesting restrictions on activity type. The most frequently selected restrictions included avoiding impact activities (93%, 235/254) and contact activities (58%, 147/254), followed by weight-bearing activities (24%, 62/254). Respondents were not aware of existing PA guidelines, although 57% expressed interest in following guidelines and 39% were uncertain. Conclusions: While there is consensus among orthopaedic surgeons that children with chronic hip conditions should engage in PA, there is considerable va

This research addresses the crucial role of the industrial sector in achieving a sustainable future amid escalating anthropogenic pressures. The study explores a Technology Center's perspective on and understanding of the future through two visual design workshops centered around the center's visualizations. The results show that, in the Technology Center, the current understanding of the visualizations in relation to the future is shaped by the idea of the visualizations as a support for making (new) things, for example new technologies, robots, systems for human-machine interactions, physical and digital spaces for industrial development. The results also show that the stories around a sustainable future evolved around economic and social sustainability, yet the narratives around the environmental aspects are less predominant. The paper contributes valuable knowledge at the intersection of sustainability, design, and industry. Keywords: visualization, sustainable future, visual design method, manufacturing

OBJECTIVES: Children with brain cancer and their families have complex care needs throughout diagnosis, active treatment, long-term survivorship, and the palliative phase of illness. This study aimed to explore the perspectives of Australian specialist clinicians on barriers and facilitators to health care for children with brain cancer and their families. METHODS: A qualitative approach was taken using semi-structured interviews. Eligible participants were clinicians of any discipline providing care to children with brain cancer and their families in Australia. Interviews were conducted by telephone and asked about perceived strengths and weaknesses in health care and available resources for this population. Qualitative content analysis used a directed approach with inductive refinement. RESULTS: Eleven clinicians participated, 5 of whom were medical, 3 nursing, and 3 allied health. The overarching theme was that the rarity and diversity of brain tumors in children confers challenges to care that lead to variation in practice. Participants reported having to adapt care from guidelines and patient/family resources designed for adults with brain cancer and children with other cancers, and rely on clinical and research networks. Specialist comprehensive cancer care was generally perceived to offer the best model for accommodating the unique needs of each child/family, but barriers to access were highlighted for children in remote Australia, and long-term follow-up was perceived to be inadequate regardless of where children lived. SIGNIFICANCE OF RESULTS: Until further brain cancer-specific paediatric guidelines become available, our findings highlight the need for communities of practice to share resources and reduce unwarranted variation. CONCLUSION: Future research should focus on developing and evaluating guidelines and other resources specific to children with brain cancer, as well as informing suitable models for long-term follow-up care for survivors.

BackgroundPediatric renal trauma is rare and lacks sufficient population-specific data to generate evidence-based management guidelines. A nonoperative approach is preferred and has been shown to be safe. However, bleeding risk assessment and management of collecting system injury are not well understood. We introduce the Multi-institutional Pediatric Acute Renal Trauma Study (Mi-PARTS), a retrospective cohort study designed to address these questions. This article describes the demographics and contemporary management of pediatric renal trauma at Level I trauma centers in the United States.MethodsRetrospective data were collected at 13 participating Level I trauma centers on pediatric patients presenting with renal trauma between 2010 and 2019. Data were gathered on demographics, injury characteristics, management, and short-term outcomes. Descriptive statistics were used to report on demographics, acute management, and outcomes.ResultsIn total, 1,216 cases were included in this study. Of all patients, 67.2% were male, and 93.8% had a blunt injury mechanism. In addition, 29.3% had isolated renal injuries, and 65.6% were high-grade (American Association for the Surgery of Trauma Grades III-V) injuries. The mean Injury Severity Score was 20.5. Most patients were managed nonoperatively (86.4%), and 3.9% had an open surgical intervention, including 2.7% having nephrectomy. Angioembolization was performed in 0.9%. Collecting system intervention was performed in 7.9%. Overall mortality was 3.3% and was only observed in patients with multiple injuries. The rate of avoidable transfer was 28.2%.ConclusionThe management and outcomes of pediatric renal trauma lack data to inform evidence-based guidelines. Nonoperative management of bleeding following renal injury is a well-established practice. Intervention for renal trauma is rare. Our findings reinforce differences from the adult population and highlights opportunities for further investigation. With data made available thr

BACKGROUND: Recurrent bacterial cystitis, often referred to as recurrent urinary tract infection (UTI), can be difficult to manage and alternative treatments are needed. HYPOTHESIS/OBJECTIVE: Intravesicular administration of asymptomatic bacteriuria (ASB) E. coli 212 will not be inferior to antimicrobial treatment for the management of recurrent UTI in dogs. ANIMALS: Thirty-four dogs with >1 UTI in the 12 months before presentation. METHODS: All dogs were deemed normal otherwise based on absence of abnormalities on physical examination, CBC, serum biochemical panel, and abdominal ultrasonography. Dogs were randomized to 1 of 2 treatment groups: Group 1 antimicrobials for 7 days or group 2 intravesicular administration of ASB E. coli 212. Owners were provided a voiding questionnaire regarding their dogs clinical signs, which was completed daily for 14 days to assess clinical cure. Dogs were examined on days 7 and 14 to assess clinical cure, and urine specimens were submitted for urinalysis and bacterial culture. RESULTS: Clinical cure rates for ASB E. coli 212-treated dogs were not inferior to 7 days of antimicrobial treatment with a 12% margin of difference to determine non-inferiority. No significant difference was found between the treatment groups on days 7 and 14 in the proportion of dogs achieving ≥50% or ≥75% reduction in their clinical score compared with baseline. CONCLUSIONS AND CLINICAL IMPORTANCE: These data suggest that intravesicular administration of ASB E. coli 212 is not inferior to antimicrobials for the treatment of recurrent UTI in dogs. This biotherapeutic agent could help alleviate the need for antimicrobials for some dogs with recurrent UTI, improving antimicrobial stewardship.

The first galaxies formed approximately 13 billion years ago, when the universe was less than a billion years old, during an epoch called the cosmic dawn. Most of these early galaxies are too distant and faint to be observed directly with our telescopes. However, their properties can, in principle, be inferred from the large-scale 21cm signal emitted by the intergalactic neutral hydrogen during cosmic dawn. Next-generation radio interferometers, such as the Square Kilometer Array (SKA), promise to provide accurate measurements of the 21cm signal. To leverage this data and infer information about the structure of our young universe, it is crucial to have fast and reliable tools to simulate the 21cm signal and efficiently explore the vast unknown astrophysical and cosmological parameter space. In this thesis, we introduce BEoRN (Bubbles during the Epoch of Reionisation Numerical simulator), a new open-source Python package designed to simulate the 21cm signal. BEoRN offers different options for the user to choose between speed and precision in its simulations. Based on the assumption that the state of the intergalactic medium around galaxies can be described by one-dimensional profiles, BEoRN efficiently generates 3D cosmological simulations of the 21cm signal within approximately 30 minutes on a single core. Additionally, the framework incorporates BEoRN-HM, an analytical approach based on the halo model, capable of predicting the 21cm power spectrum in just a few tens of seconds. With these new tools, we explore the potential of the 21cm signal to reveal the properties of the first galaxies, which, for most of them, are invisible to our telescopes. At the map level, we show the variety of different morphologies of the 21cm signal still allowed by current observations, providing convincing evidence that the properties of the early galaxies are readable in the pattern of 21cm maps. At the quantitative level, we use BEoRN-HM to conduct a cosmological inference on mock

Geologically, the Arctic is one of the least-explored regions of Earth. Obtaining data in the high Arctic is logistically, economically, and environmentally expensive, but the township of Longyearbyen (population of 2617 as of 2024) at 78° N represents a relatively easily accessible gateway to Arctic geology and is home to The University Centre in Svalbard (UNIS). These unique factors provide a foundation from which to teach and explore Arctic geology via the classroom, the laboratory, and the field. UNIS was founded in 1993 as the Norwegian “field university”, offering field-based courses in Arctic geology, geophysics, biology, and technology to students from Norway and abroad. In this contribution, we present one of the educational components of the international collaboration project NOR-R-AM (a Norwegian-Russian-North American collaboration in Arctic research and collaboration, titled Changes at the Top of the World through Volcanism and Plate Tectonics) which ran from 2017 to 2024. One of the key deliverables of NOR-R-AM was a new graduate (Master's and PhD-level) course called Arctic Tectonics and Volcanism that we have established and taught annually at UNIS since 2018 and detail herein. The course's main objective is to teach the complex geological evolution of the Arctic from the Devonian period (∼ 420 million years ago, Ma) to the present day through integrating multi-scale datasets and a broad range of geoscientific disciplines. We outline the course itself before presenting student perspectives based on both an anonymous questionnaire (n=27) and in-depth perceptions of four selected students. The course, with an annual intake of up to 20 MSc and PhD students, is held over a 6-week period, typically in spring or autumn. The course comprises modules on field and polar safety, Svalbard/Barents Sea geology, wider Arctic geology, plate tectonics, mantle dynamics, geo- and thermochronology, and geochemistry of igneous systems. A field component, which in some

Scenarios have been an important integrating element in the Sixth Assessment Report (AR6) of the Intergovernmental Panel on Climate Change (IPCC) in the understanding of possible climate outcomes, impacts and risks, and mitigation futures. Integration supports a consistent, coherent assessment, new insights and the opportunity to address policy-relevant questions that would not be possible otherwise, for example, which impacts are unavoidable, which are reversible, what is a consistent remaining carbon budget to keep temperatures below a level and what would be a consistent route of action to achieve that goal. The AR6 builds on community frameworks that are developed to support a coherent use of scenarios across the assessment, yet their use in the assessment and the related timelines presented coordination challenges. From lessons within each Working Group (WG) assessment and the cross-WG experience, we present insights into the role of scenarios in future assessments, including the enhanced integration of impacts into scenarios, near-term information and community coordination efforts. Recommendations and opportunities are discussed for how scenarios can support strengthened consistency and policy relevance in the next IPCC assessment cycle.

Background: Endotracheal intubation (ETI) in the intensive care unit (ICU) poses a significantly higher risk of life-threatening complications compared to procedures in the operating room (OR). Limited resources and the complex, critical condition of ICU patients make ETI more challenging. To bridge the gap in care, implementing standardized checklists and simulation training for ETI procedures in intensive care units will enhance patient outcomes by ensuring proper preparation, equipment, medications, and communication between healthcare providers. Purpose: This project aims to improve knowledge and ICU nurses’ confidence in preparing for and managing endotracheal intubation procedures in critically ill patients, thereby enhancing patient safety during emergent intubations outside the operating room. Methods: The Lewin’s Three-Step Model of Change and the Iowa model of evidence-based practice provided context for implementation. A PowerPoint presentation which reviewed the contents of a standardized preparedness checklist was followed by a simulation exercise led by the primary investigator. A mixed-methods design using a pre-and post-test survey was utilized for data collection. Results: Sixty-two ICU RNs participated in the intervention. A paired T-test was utilized to analyze the data from the pre-and post-test surveys. 70% of ICU RNs participating in this project reported increased confidence with the greatest improvement in self-reported confidence exemplified in the group of nurses with less than two years of experience working in the ICU. Conclusion: The results indicated that the intervention effectively increased confidence and knowledge of ETI, which improves patient safety and decreases morbidity and mortality. Additionally, results indicated that simulation training is beneficial to increase confidence levels while preparing for life-threatening emergencies in a safe, controlled environment.

INTRODUCTION: Brace treatment is common to address radiological dysplasia in infants with developmental dysplasia of the hip (DDH); however, it is unclear whether bracing provides significant benefit above careful observation by ultrasound. If observation alone is non-inferior to bracing for radiological dysplasia, unnecessary treatment may be avoided. Therefore, the purpose of this study is to determine whether observation is non-inferior to bracing for infants with radiological dysplasia. METHODS AND ANALYSIS: This will be a multicentre, global, randomised, non-inferiority trial performed under the auspices of a global prospective registry for infants and children diagnosed with DDH. Patients will be included if they present with radiological dysplasia (centred hip, alpha angle 43-60°, percent femoral head coverage greater than 35% measured on ultrasound) of a clinically stable hip under 3 months old. Patients will be excluded if they present with clinical hip instability, have received prior treatment or have known/suspected neuromuscular, collagen, chromosomal or lower-extremity congenital abnormalities or syndromic-associated hip abnormalities. Patients will be enrolled and randomised to undergo observation alone or brace treatment with a Pavlik harness for a minimum of 6 weeks. Follow-up visits will occur at 6 weeks, 1 year and 2 years post-enrolment. The primary outcome will be the norm-referenced acetabular index measured on the 2-year radiograph with a 3° non-inferiority margin. A total of 514 patients will be included.The study is anticipated to start in April 2024 and end in September 2028.The primary outcome will be compared between arms with a mixed-effects model with a random intercept for study centre, and a single covariate for the treatment group. If the lower bound of the 95% CI lies within 3° of the mean, we will treat this as evidence for non-inferiority. ETHICS AND DISSEMINATION: Ethics approval has been obtained from the lead sites ethics board

Vaccination against the novel SARS-CoV-2 virus was highly effective in the general population at protecting against infection, however for immunocompromised individuals the mRNA vaccines alone were not sufficient in providing an immune response to protect against the virus. The inadequate response to vaccination indicated the need for a pre-exposure prophylaxis agent to provide immunocompromised individuals an alternate form of passive immunity to the virus, further decreasing the risk of developing a severe response to the virus. Monoclonal antibodies (mAbs) were developed and tested for use as a prevention method, illustrating increased efficacy among immunocompromised populations where a sharp decrease in severe clinical adverse events (death, hospitalization) was noted amongst recipients. Given the severity of SARS-CoV-2 infection amongst immunocompromised patients, where the virus has a higher degree of severity and often a prolonged course of disease, there is an increased need to investigate current mAbs in terms of pre-exposure prophylaxis effectiveness and general efficacy of mAbs as prophylaxis are relatively new and many gaps in current literature exist. The importance of effective mAb pre-exposure prophylaxis for immunocompromised patients highlights the scope of Public Health objectives, specifically in aiding those who feel disease risk more strongly.

Climate change is one of the biggest threats to global food security, with observed impacts already affecting agriculture. This study aims to systematize and analyze the observed biophysical impacts and their evolution in agriculture in Brazil. For this, we carry out a systematic literature review that includes 1844 articles in the first stage, and 53 articles with synthesized information retrieved. Temperature and precipitation are the most studied climate variables when considering observed climate impact on food production, with corn, soybeans, and sugarcane as the major crops assessed. We also identify regional patterns of both positive and negative trends due to climate change: 67% of assessed studies report negative impacts, 15% both negative and positive, 11% neutral relationships and only 7% reveal positive effects. The diversity in scope and methodological approaches across studies presents a challenge, as only a fraction sufficiently contextualizes baseline conditions, not allowing for a comprehensive understanding of impact attribution. Nonetheless, the literature spotlights productivity declines in cassava, cocoa, rice and wheat. As for corn, soybeans, and sugarcane studies reveal both positive and negative impacts, depending on baseline premises. The call for more transparent and comprehensive studies is urgent, especially to encompass a broader range of food crops, particularly in family farming systems and across diverse regional scales. Such studies are imperative for advancing evidence-based climate-resilient strategies in agriculture aiming to map and prevent negative impacts, while promoting positive outcomes in food production.

The consensus-based guideline "hand antisepsis and hand hygiene" for Germany has the following sections: Prevention of nosocomial infections by hygienic hand antisepsis, prevention of surgical site infections by surgical hand antisepsis, infection prevention in the community by hand antisepsis in epidemic or pandemic situations, hand washing, selection of alcohol-based hand rubs and wash lotions, medical gloves and protective gloves, preconditions for hand hygiene, skin protection and skin care, quality assurance of the implementation of hand hygiene measures and legal aspects. The guideline was developed by the German Society for Hospital Hygiene in cooperation with 22 professional societies, 2 professional organizations, the German Care Council, the Federal Working Group for Self-Help of People with Disabilities and Chronic Illness and their Family Members, the General Accident Insurance Institution Austria and the German-speaking Interest Group of Infection Prevention Experts and Hospital Hygiene Consultants., Die Konsensus basierte Leitlinie ist in folgende Themen untergliedert: Prävention nosokomialer Infektionen durch hygienische Händeantiseptik, Prävention postoperativer Wundinfektionen durch chirurgische Händeantiseptik, Infektionsprävention in der Bevölkerung durch Händeantiseptik bei epidemischer bzw. pandemischer Situation, Händewaschung aus sozialer und infektionspräventiver Indikation, Auswahl von Alkohol-basierten Händeantiseptika und von Waschlotionen, medizinische Handschuhe und Schutzhandschuhe, Voraussetzungen für die Händehygiene, Hautschutz und Hautpflege, Qualitätssicherung der Umsetzung der Maßnahmen der Händehygiene und rechtliche Aspekte. Die Leitlinie wurde von der Deutschen Gesellschaft für Krankenhaushygiene in Zusammenarbeit mit 22 Fachgesellschaften, 2 Berufsverbänden, dem Deutschen Pflegerat, der Bundesarbeitsgemeinschaft Selbsthilfe von Menschen mit Behinderung und chronischer Erkrankung und ihren Angehörigen, der Allgemeinen Unfallversicherungsanstalt Österreich und der Deutschprachigen Interessengruppe der Fachexperten/innen für Infektionsprävention und Berater für Spitalhygiene erarbeitet.

During the epoch of reionization (EoR), the 21-cm signal allows direct observation of the neutral hydrogen (HI) in the intergalactic medium (IGM). In the post-reionization era, this signal instead probes HI in galaxies, which traces the dark matter density distribution. With new numerical simulations, we investigated the end stages of reionization to elucidate the transition of our Universe into the post-reionization era. Our models are consistent with the latest high-redshift measurements, including ultraviolet (UV) luminosity functions up to redshift similar or equal to 8. Notably, these models consistently reproduced the evolution of the UV photon background, which is constrained from Lyman-alpha absorption spectra. We studied the dependence of this background on the nature of photon sinks in the IGM, requiring mean free path of UV photons to be similar to 10 comoving-megaparsecs (cMpc) during the EoR that increases gradually with time during late stages (z less than or similar to 6). Our models revealed that the reionization of the IGM transitioned from an inside-out to an outside-in process when the Universe is less than 0.01 per cent neutral. During this epoch, the 21-cm signal also shifted from probing predominantly the H I in the IGM to that in galaxies. Furthermore, we identified a statistically significant number of large neutral islands (with sizes up to 40 cMpc) persisting until very late stages (5 less than or similar to z less than or similar to 6) that can imprint features in Lyman-alpha absorption spectra and also produce a knee-like feature in the 21-cm power spectrum., QC 20240912

Predicting the 21-cm signal from the epoch of reionization and cosmic dawn is a complex and challenging task. Various simplifying assumptions have been applied over the last decades to make the modeling more affordable. In this paper, we investigate the validity of several such assumptions, using a simulation suite consisting of three different astrophysical source models that agree with the current constraints on the reionization history and the UV luminosity function. We first show that the common assumption of a saturated spin temperature may lead to significant errors in the 21-cm clustering signal over the full reionization period. The same is true for the assumption of a neutral universe during the cosmic dawn which may lead to significant deviation from the correct signal during the heating and the Lyman-alpha coupling period. Another popular simplifying assumption consists of predicting the global differential brightness temperature (dTb) based on the average quantities of the reionization fraction, gas temperature, and Lyman-alpha coupling. We show that such an approach leads to a 10 percent deeper absorption signal compared to the results obtained by averaging the final dTb-map. Finally, we investigate the simplifying method of breaking the 21-cm clustering signal into different auto and cross components that are then solved assuming linearity. We show that even though the individual fields have a variance well below unity, they often cannot be treated perturbatively as the perturbations are strongly non-Gaussian. As a consequence, predictions based on the perturbative solution of individual auto and cross power spectra may lead to strongly biased results, even if higher-order terms are taken into account., QC 20240822

Erworben im Rahmen der Schweizer Nationallizenzen (http://www.nationallizenzen.ch), Eine wesentliche Voraussetzung für eine erfolgreiche digitale Transformation des Gesundheitswesens ist eine gut ausgeprägte digitale Gesundheitskompetenz (DGK) der Bevölkerung. DGK ist die Fähigkeit zum Umgang mit gesundheitsbezogenen digitalen Informationen und Informationsmöglichkeiten mit dem Ziel, Gesundheit und Wohlbefinden für sich selbst und sein Umfeld zu fördern und zu erhalten. Der Artikel beleuchtet die Diskussion über DGK, vorhandene Studien und die darin verwendeten Messinstrumente sowie die Datenlage in Deutschland und erörtert aktuelle Herausforderungen. DGK besteht aus verschiedenen Teilkompetenzen, die aktuelle digitale Informationsverhalten, -möglichkeiten und -risiken widerspiegeln. Die Datenlage ist, aufgrund unterschiedlicher Studiendesigns und -instrumente, sehr heterogen, was die Aussagekraft limitiert. Zwei repräsentative Studien, HLS-GER 2 der Universität Bielefeld sowie die Studie der AOK Rheinland/Hamburg und des Leibniz-WissenschaftsCampus, weisen trotz unterschiedlicher Methoden auf einen hohen Anteil von Menschen mit geringer DGK hin. National wie international zeigt sich, dass die DGK einem sozialen Gradienten unterliegt und mit Bildungsniveau, Sozialstatus, finanzieller Deprivation und Alter assoziiert ist. Die DGK ist in Deutschland den vorliegenden Daten zufolge noch unzureichend; somit besteht ein großer Handlungsbedarf. Erforderliche gesetzliche Rahmenbedingungen sind gegeben, dennoch fehlt es an verlässlichen finanziellen Ressourcen ebenso an einer soliden Datengrundlage auf Bevölkerungsebene zu DGK. Damit ließen sich Vulnerabilitätsfaktoren identifizieren und die Implementation von Maßnahmen vorbereiten und evaluieren. Zudem bedarf es einer vertiefenden konzeptionellen Diskussion zur DGK, die an das etablierte Gesundheitskompetenzkonzept anknüpft und auch die gesundheitsbezogene Infodemie und ihre Folgen für die DGK aufgreift.

The authors are all devoted energy system and sustainability transformation scholars, who collaborate regularly and actively at global and local levels to advance the knowledge space of demand-side solutions and policies. They are members of a growing bottom-up initiative, the Energy Demand Changes Induced by Technological and Social Innovations (EDITS) network (https://iiasa.ac.at/projects/edits), which builds on various research disciplines to facilitate advances in modeling, data compilation, and analysis of the scope and breadth of the potential contributions of demand-side solutions for climate change mitigation, improved wellbeing for all, and sustainability, complementing supply-side solutions for decarbonizing the energy and material systems.

[Background] Pathogenic variants in several genes have been linked to genetic forms of isolated or combined dystonia. The phenotypic and genetic spectrum and the frequency of pathogenic variants in these genes have not yet been fully elucidated, neither in patients with dystonia nor with other, sometimes co-occurring movement disorders such as Parkinson's disease (PD)., [Objectives] To screen >2000 patients with dystonia or PD for rare variants in known dystonia-causing genes., [Methods] We screened 1207 dystonia patients from Germany (DysTract consortium), Spain, and South Korea, and 1036 PD patients from Germany for pathogenic variants using a next-generation sequencing gene panel. The impact on DNA methylation of KMT2B variants was evaluated by analyzing the gene's characteristic episignature., [Results] We identified 171 carriers (109 with dystonia [9.0%]; 62 with PD [6.0%]) of 131 rare variants (minor allele frequency <0.005). A total of 52 patients (48 dystonia [4.0%]; four PD [0.4%, all with GCH1 variants]) carried 33 different (likely) pathogenic variants, of which 17 were not previously reported. Pathogenic biallelic variants in PRKRA were not found. Episignature analysis of 48 KMT2B variants revealed that only two of these should be considered (likely) pathogenic., [Conclusion] This study confirms pathogenic variants in GCH1, GNAL, KMT2B, SGCE, THAP1, and TOR1A as relevant causes in dystonia and expands the mutational spectrum. Of note, likely pathogenic variants only in GCH1 were also found among PD patients. For DYT-KMT2B, the recently described episignature served as a reliable readout to determine the functional effect of newly identified variants.

Predicting the 21cm signal from the epoch of reionization and cosmic dawn is a complex and challenging task. Various simplifying assumptions have been applied over the last decades to make the modeling more affordable. In this paper, we investigate the validity of several such assumptions, using a simulation suite consisting of three different astrophysical source models that agree with the current constraints on the reionization history and the UV luminosity function. We first show that the common assumption of a saturated spin temperature may lead to significant errors in the 21cm clustering signal over the full reionization period. The same is true for the assumption of a neutral universe during the cosmic dawn which may lead to significant deviation from the correct signal during the heating and the Lyman-$\alpha$ coupling period. Another popular simplifying assumption consists of predicting the global differential brightness temperature ($dT_b$) based on the average quantities of the reionization fraction, gas temperature, and Lyman-$\alpha$ coupling. We show that such an approach leads to a 10 percent deeper absorption signal compared to the results obtained by averaging the final $dT_b$-map. Finally, we investigate the simplifying method of breaking the 21cm clustering signal into different auto and cross components that are then solved assuming linearity. We show that even though the individual fields have a variance well below unity, they often cannot be treated perturbatively as the perturbations are strongly non-Gaussian. As a consequence, predictions based on the perturbative solution of individual auto and cross power spectra may lead to strongly biased results, even if higher-order terms are taken into account.

We present results from X-ray imaging of high-aspect-ratio magnetic reconnection experiments driven at the National Ignition Facility. Two parallel, self-magnetized, elongated laser-driven plumes are produced by tiling 40 laser beams. A magnetic reconnection layer is formed by the collision of the plumes. A gated X-ray framing pinhole camera with micro-channel plate (MCP) detector produces multiple images through various filters of the formation and evolution of both the plumes and current sheet. As the diagnostic integrates plasma self-emission along the line of sight, 2-dimensional electron temperature maps $\langle T_e \rangle_Y$ are constructed by taking the ratio of intensity of these images obtained with different filters. The plumes have a characteristic temperature $\langle T_e \rangle_Y = 240 \pm 20$ eV at 2 ns after the initial laser irradiation and exhibit a slow cooling up to 4 ns. The reconnection layer forms at 3 ns with a temperature $\langle T_e \rangle_Y = 280 \pm 50$ eV as the result of the collision of the plumes. The error bars of the plumes and current sheet temperatures separate at $4$ ns, showing the heating of the current sheet from colder inflows. Using a semi-analytical model, we find that the observed heating of the current sheet is consistent with being produced by electron-ion drag, rather than the conversion of magnetic to kinetic energy., Comment: Submitted to Physics of Plasmas. 19 pages (total), 14 figures, 2 tables

Mini-magnetospheres are small ion-scale structures that are well-suited to studying kinetic-scale physics of collisionless space plasmas. Such ion-scale magnetospheres can be found on local regions of the Moon, associated with the lunar crustal magnetic field. In this paper, we report on the laboratory experimental study of magnetic reconnection in laser-driven, lunar-like ion-scale magnetospheres on the Large Plasma Device (LAPD) at the University of California - Los Angeles. In the experiment, a high-repetition rate (1 Hz), nanosecond laser is used to drive a fast moving, collisionless plasma that expands into the field generated by a pulsed magnetic dipole embedded into a background plasma and magnetic field. The high-repetition rate enables the acquisition of time-resolved volumetric data of the magnetic and electric fields to characterize magnetic reconnection and calculate the reconnection rate. We notably observe the formation of Hall fields associated with reconnection. Particle-in-cell simulations reproducing the experimental results were performed to study the micro-physics of the interaction. By analyzing the generalized Ohm's law terms, we find that the electron-only reconnection is driven by kinetic effects, through the electron pressure anisotropy. These results are compared to recent satellite measurements that found evidence of magnetic reconnection near the lunar surface.

Associative memory and probabilistic modeling are two fundamental topics in artificial intelligence. The first studies recurrent neural networks designed to denoise, complete and retrieve data, whereas the second studies learning and sampling from probability distributions. Based on the observation that associative memory's energy functions can be seen as probabilistic modeling's negative log likelihoods, we build a bridge between the two that enables useful flow of ideas in both directions. We showcase four examples: First, we propose new energy-based models that flexibly adapt their energy functions to new in-context datasets, an approach we term \textit{in-context learning of energy functions}. Second, we propose two new associative memory models: one that dynamically creates new memories as necessitated by the training data using Bayesian nonparametrics, and another that explicitly computes proportional memory assignments using the evidence lower bound. Third, using tools from associative memory, we analytically and numerically characterize the memory capacity of Gaussian kernel density estimators, a widespread tool in probababilistic modeling. Fourth, we study a widespread implementation choice in transformers -- normalization followed by self attention -- to show it performs clustering on the hypersphere. Altogether, this work urges further exchange of useful ideas between these two continents of artificial intelligence.

Let $\alpha$ and $\beta$ belong to the same quadratic field. We show that the inhomogeneous Beatty sequence $(\lfloor n \alpha + \beta \rfloor)_{n \geq 1}$ is synchronized, in the sense that there is a finite automaton that takes as input the Ostrowski representations of $n$ and $y$ in parallel, and accepts if and only if $y = \lfloor n \alpha + \beta \rfloor$. Since it is already known that the addition relation is computable for Ostrowski representations based on a quadratic number, a consequence is a new and rather simple proof that the first-order logical theory of these sequences with addition is decidable. The decision procedure is easily implemented in the free software Walnut. As an application, we show that for each $r \geq 1$ it is decidable whether the set $\{ \lfloor n \alpha + \beta \rfloor \, : \, n \geq 1 \}$ forms an additive basis (or asymptotic additive basis) of order $r$. Using our techniques, we also solve some open problems of Reble and Kimberling, and give an explicit characterization of a sequence of Hildebrand et al.

Language models pre-trained on web-scale corpora demonstrate impressive capabilities on diverse downstream tasks. However, there is increasing concern whether such capabilities might arise from evaluation datasets being included in the pre-training corpus -- a phenomenon known as \textit{data contamination} -- in a manner that artificially increases performance. There has been little understanding of how this potential contamination might influence LMs' performance on downstream tasks. In this paper, we explore the impact of data contamination at the pre-training stage by pre-training a series of GPT-2 models \textit{from scratch}. We highlight the effect of both text contamination (\textit{i.e.}\ input text of the evaluation samples) and ground-truth contamination (\textit{i.e.}\ the prompts asked on the input and the desired outputs) from evaluation data. We also investigate the effects of repeating contamination for various downstream tasks. Additionally, we examine the prevailing n-gram-based definitions of contamination within current LLM reports, pinpointing their limitations and inadequacy. Our findings offer new insights into data contamination's effects on language model capabilities and underscore the need for independent, comprehensive contamination assessments in LLM studies., Comment: 16 pages, 5 figures

Deliberately removing carbon dioxide from the atmosphere is an important element of bringing mitigation pathways in line with the climate goals of the Paris Agreement. To reach global net-zero CO2 emissions and limit global warming to 1.5°C with no or limited overshoot, global mitigation pathways assessed by IPCC’s Sixth Assessment Report require some world regions to achieve net-negative CO2 emissions with large-scale carbon dioxide removal (CDR) deployment. This raises important questions about the availability and feasibility of CDR deployment in different societal and political contexts. This paper therefore combines an analysis of CDR deployment in a sample of scenarios from the IPCC AR6 database with a bottom-up analysis of the state of CDR governance and policy in countries considered key in scaling up CDR capacity and not yet covered by existing research. In particular, the paper focuses on Brazil, China, and India as important emerging economies and large emitters. We highlight the expected use of CDR methods in those regions in scenarios and systematically assess and compare the level of CDR regulation and innovation across these countries. This comparative perspective has the potential to broaden the understanding of existing and emerging CDR policies and politics. The synthesis of the case studies provides three key contributions to existing literature: First, we explore the state of CDR governance and policymaking in key emerging economies. As in OECD countries, there is a notable lack of CDR regulation and innovation to enable the scale of CDR required in the short- and medium term. Second, we identify that repurposing policies is a key type of emerging CDR policymaking in these countries targeting CDR methods in the land use, land use change and forestry (LULUCF) sector. We find that the repurposing efforts strengthen the level of regulation and innovation for this group of methods. Third, we explore three building blocks (regional differentiation, de

In recent years, numerous embedding models have been made available and widely used for various NLP tasks. Choosing a model that performs well for several tasks in English has been largely simplified by the Massive Text Embedding Benchmark (MTEB), but extensions to other languages remain challenging. This is why we expand MTEB to propose the first massive benchmark of sentence embeddings for French. Not only we gather 22 existing datasets in an easy-to-use interface, but we also create three new French datasets for a global evaluation over 8 different tasks. We perform a large scale comparison with 46 carefully selected embedding models, conduct comprehensive statistical tests, and analyze the correlation between model performance and many of their characteristics. We find out that even if no model is the best on all tasks, large multilingual models pre-trained on sentence similarity perform particularly well. Our work comes with open-source code, new datasets and a public leaderboard.

Given many copies of an unknown quantum state $\rho$, we consider the task of learning a classical description of its principal eigenstate. Namely, assuming that $\rho$ has an eigenstate $|\phi\rangle$ with (unknown) eigenvalue $\lambda > 1/2$, the goal is to learn a (classical shadows style) classical description of $|\phi\rangle$ which can later be used to estimate expectation values $\langle \phi |O| \phi \rangle$ for any $O$ in some class of observables. We consider the sample-complexity setting in which generating a copy of $\rho$ is expensive, but joint measurements on many copies of the state are possible. We present a protocol for this task scaling with the principal eigenvalue $\lambda$ and show that it is optimal within a space of natural approaches, e.g., applying quantum state purification followed by a single-copy classical shadows scheme. Furthermore, when $\lambda$ is sufficiently close to $1$, the performance of our algorithm is optimal--matching the sample complexity for pure state classical shadows., Comment: 38 pages

Network programmability allows modification of fine-grain data plane functionality. The performance benefits of data plane programmability have motivated many researchers to offload computation that previously operated only on servers to the network, creating the notion of in-network computing (INC). Because failures can occur in the data plane, fault tolerance mechanisms are essential for INC. However, INC operators and developers must manually set fault tolerance requirements using domain knowledge to change the source code. These manually set requirements may take time and lead to errors in case of misconfiguration. In this work, we present Araucaria, a system that aims to simplify the definition and implementation of fault tolerance requirements for INC. The system allows requirements specification using an intent language, which enables the expression of consistency and availability requirements in a constrained natural language. A refinement process translates the intent and incorporates the essential building blocks and configurations into the INC code. We present a prototype of Araucaria and analyze the end-to-end system behavior. Experiments demonstrate that the refinement scales to multiple intents and that the system provides fault tolerance with negligible overhead in failure scenarios.

Tuning parameters in model predictive control (MPC) presents significant challenges, particularly when there is a notable discrepancy between the controller's predictions and the actual behavior of the closed-loop plant. This mismatch may stem from factors like substantial model-plant differences, limited prediction horizons that do not cover the entire time of interest, or unforeseen system disturbances. Such mismatches can jeopardize both performance and safety, including constraint satisfaction. Traditional methods address this issue by modifying the finite horizon cost function to better reflect the overall operational cost, learning parts of the prediction model from data, or implementing robust MPC strategies, which might be either computationally intensive or overly cautious. As an alternative, directly optimizing or learning the controller parameters to enhance closed-loop performance has been proposed. We apply Bayesian optimization for efficient learning of unknown model parameters and parameterized constraint backoff terms, aiming to improve closed-loop performance of battery fast charging. This approach establishes a hierarchical control framework where Bayesian optimization directly fine-tunes closed-loop behavior towards a global and long-term objective, while MPC handles lower-level, short-term control tasks. For lithium-ion battery fast charging, we show that the learning approach not only ensures safe operation but also maximizes closed-loop performance. This includes maintaining the battery's operation below its maximum terminal voltage and reducing charging times, all achieved using a standard nominal MPC model with a short horizon and notable initial model-plant mismatch., Comment: 6 pages, 5 figures, accepted for ADCHEM 2024

The proliferation of generative models, combined with pretraining on web-scale data, raises a timely question: what happens when these models are trained on their own generated outputs? Recent investigations into model-data feedback loops proposed that such loops would lead to a phenomenon termed model collapse, under which performance progressively degrades with each model-data feedback iteration until fitted models become useless. However, those studies largely assumed that new data replace old data over time, where an arguably more realistic assumption is that data accumulate over time. In this paper, we ask: what effect does accumulating data have on model collapse? We empirically study this question by pretraining sequences of language models on text corpora. We confirm that replacing the original real data by each generation's synthetic data does indeed tend towards model collapse, then demonstrate that accumulating the successive generations of synthetic data alongside the original real data avoids model collapse; these results hold across a range of model sizes, architectures, and hyperparameters. We obtain similar results for deep generative models on other types of real data: diffusion models for molecule conformation generation and variational autoencoders for image generation. To understand why accumulating data can avoid model collapse, we use an analytically tractable framework introduced by prior work in which a sequence of linear models are fit to the previous models' outputs. Previous work used this framework to show that if data are replaced, the test error increases with the number of model-fitting iterations; we extend this argument to prove that if data instead accumulate, the test error has a finite upper bound independent of the number of iterations, meaning model collapse no longer occurs.