Diffusion models, and their generalization, flow matching, have had a remarkable impact on the field of media generation. Here, the conventional approach is to learn the complex mapping from a simple source distribution of Gaussian noise to the target media distribution. For cross-modal tasks such as text-to-image generation, this same mapping from noise to image is learnt whilst including a conditioning mechanism in the model. One key and thus far relatively unexplored feature of flow matching is that, unlike Diffusion models, they are not constrained for the source distribution to be noise. Hence, in this paper, we propose a paradigm shift, and ask the question of whether we can instead train flow matching models to learn a direct mapping from the distribution of one modality to the distribution of another, thus obviating the need for both the noise distribution and conditioning mechanism. We present a general and simple framework, CrossFlow, for cross-modal flow matching. We show the importance of applying Variational Encoders to the input data, and introduce a method to enable Classifier-free guidance. Surprisingly, for text-to-image, CrossFlow with a vanilla transformer without cross attention slightly outperforms standard flow matching, and we show that it scales better with training steps and model size, while also allowing for interesting latent arithmetic which results in semantically meaningful edits in the output space. To demonstrate the generalizability of our approach, we also show that CrossFlow is on par with or outperforms the state-of-the-art for various cross-modal / intra-modal mapping tasks, viz. image captioning, depth estimation, and image super-resolution. We hope this paper contributes to accelerating progress in cross-modal media generation.

BACKGROUND:Many patients receive repeat prescriptions for routine medications used to treat chronic conditions. Doctors typically issue repeat prescriptions with durations ranging from 28 to 84 days. There is currently no national guidance in England for the optimal prescription duration for routine medications. AIM:To evaluate current prescription durations for five common routine medications in England; explore and visualise geographical variation; and identify practice factors that are associated with shorter prescribing duration to inform policy making. DESIGN AND SETTING:A retrospective cohort study of NHS primary care prescribing data in England from December 2018 to November 2019. METHOD:The prescription duration was analysed for five common routine medications in England; ramipril, atorvastatin, simvastatin, levothyroxine, and amlodipine. Variation was assessed between regional clinical commissioning groups (CCGs), and practice factors associated with different durations were identified. RESULTS:Of the common medications included, 28-day prescriptions accounted for 48.5% (2.5 billion) tablets/capsules issued, while 43.6% were issued for 56 days. There was very wide regional variation (7.2%-95.0%) in the proportion of 28-day prescriptions issued by CCGs. Practice dispensing status was the most likely predictor of prescription duration; dispensing practices had a higher 28-day prescribing proportion than non-dispensing practices. The proportion of patients with chronic conditions and the electronic health record system used by a practice were also associated with prescription duration. CONCLUSION:This analysis of OpenPrescribing data showed that repeat prescriptions of 28 days are common for patients taking routine medications for chronic conditions, particularly in dispensing practices. This provides data to inform the policy debate on current practice. Configuration of electronic health record systems offer an opportunity to implement and evaluate new policies on repeat prescription duration in England.

Background COVID-19 lockdowns led to increased reports of depressive symptoms in the general population and impacted the health and social care services of people with learning disability and autism. We explored whether the COVID-19 pandemic had an impact on antidepressant prescribing trends within these and the general population. Methods With the approval of NHS England, we used >24 million patients primary care data from the OpenSAFELY-TPP platform. We identified patients with learning disability or autism and used an interrupted time series analysis to quantify trends in those prescribed and newly prescribed an antidepressant across key demographic and clinical subgroups, comparing pre-COVID-19 (January 2018-February 2020), COVID-19 lockdown (March 2020-February 2021) and the recovery period (March 2021-December 2022). Results Prior to COVID-19 lockdown, antidepressant prescribing was increasing at 0.3% (95% CI 0.2% to 0.3%) patients per month, in the general population and in those with learning disability, and 0.3% (95% CI 0.2% to 0.4%) in those with autism. We did not find evidence that the pandemic was associated with a change in trend of antidepressant prescribing in the general population (RR 1.00 (95% CI 0.97 to 1.02)), in those with autism (RR 0.99 (95% CI 0.97 to 1.01)), or in those with learning disability (RR 0.98 (95% CI 0.96 to 1.00)). New prescribing post lockdown was 13% and 12% below expected if COVID-19 had not happened in both the general population and those with autism (RR 0.87 (95% CI 0.83 to 0.93), RR 0.88 (95% CI 0.83 to 0.92))), but not learning disability (RR 0.96 (95% CI 0.87 to 1.05)). Conclusions and Implications Pre-COVID-19, antidepressant prescribing was increasing at 0.3% per month. While we did not see an impact of COVID-19 on overall prescribing in the general population, prescriptions to those aged 0-19, 20-29, and new prescriptions were lower than pre-COVID-19 trends would have predicted, but tricyclics and new prescriptions in care homes were higher than expected. ### Competing Interest Statement The authors have declared no competing interest. ### Funding Statement The OpenSAFELY Platform is supported by grants from the Wellcome Trust (222097/Z/20/Z) and MRC (MR/V015737/1, MC\_PC\_20059, MR/W016729/1). In addition, development of OpenSAFELY has been funded by the Longitudinal Health and Wellbeing strand of the National Core Studies programme (MC\_PC\_20030: MC\_PC\_20059), the NIHR funded CONVALESCENCE programme (COV-LT-0009), NIHR (NIHR135559, COV-LT2-0073), and the Data and Connectivity National Core Study funded by UK Research and Innovation (MC\_PC\_20058) and Health Data Research UK (HDRUK2021.000). BG has also received funding from: the Bennett Foundation, the Wellcome Trust, NIHR Oxford Biomedical Research Centre, NIHR Applied Research Collaboration Oxford and Thames Valley, the Mohn-Westlake Foundation; all Bennett Institute staff are supported by BGs grants on this work. BMK is also employed by NHS England working on medicines policy and clinical lead for primary care medicines data. ### Author Declarations I confirm all relevant ethical guidelines have been followed, and any necessary IRB and/or ethics committee approvals have been obtained. Yes The details of the IRB/oversight body that provided approval or exemption for the research described are given below: London - City & East Research Ethics Committee gave ethical approval for this work. REC reference 20/LO/0651. I confirm that all necessary patient/participant consent has been obtained and the appropriate institutional forms have been archived, and that any patient/participant/sample identifiers included were not known to anyone (e.g., hospital staff, patients or participants themselves) outside the research group so cannot be used to identify individuals. Yes I understand that all clinical trials and any other prospective interventional studies must be registered with an ICMJE-approved registry, such as ClinicalTrials.gov. I confirm that any such study reported in the manuscript has been registered and the trial registration ID is provided (note: if posting a prospective study registered retrospectively, please provide a statement in the trial ID field explaining why the study was not registered in advance). Yes I have followed all appropriate research reporting guidelines, such as any relevant EQUATOR Network research reporting checklist(s) and other pertinent material, if applicable. Yes The dataset analysed within OpenSAFELY is based on > 24 million people currently registered with GP surgeries using TPP SystmOne software. All data were linked, stored and analysed securely using the OpenSAFELY platform, https://www.opensafely.org/, as part of the NHS England OpenSAFELY COVID-19 service. Data include pseudonymised data such as coded diagnoses, medications and physiological parameters. No free text data are included. All code is shared openly for review and re-use under MIT open licence. Detailed pseudonymised patient data is potentially re-identifiable and therefore not shared. Data management and analysis was performed using Python 3. All code for data management and analysis, as well as codelists, is shared openly for inspection and re-use.

The VoxCeleb Speaker Recognition Challenges (VoxSRC) were a series of challenges and workshops that ran annually from 2019 to 2023. The challenges primarily evaluated the tasks of speaker recognition and diarisation under various settings including: closed and open training data; as well as supervised, self-supervised, and semi-supervised training for domain adaptation. The challenges also provided publicly available training and evaluation datasets for each task and setting, with new test sets released each year. In this paper, we provide a review of these challenges that covers: what they explored; the methods developed by the challenge participants and how these evolved; and also the current state of the field for speaker verification and diarisation.We chart the progress in performance over the five installments of the challenge on a common evaluation dataset and provide a detailed analysis of how each year's special focus affected participants' performance. This paper is aimed both at researchers who want an overview of the speaker recognition and diarisation field, and also at challenge organisers who want to benefit from the successes and avoid the mistakes of the VoxSRC challenges.We end with a discussion of the current strengths of the field and open challenges.

AimsThe COVID‐19 pandemic caused significant disruption to routine activity in primary care. Medication reviews are an important primary care activity ensuring safety and appropriateness of prescribing. A disruption could have significant negative implications for patient care. Using routinely collected data, our aim was first to describe codes used to record medication review activity and then to report the impact of COVID‐19 on the rates of medication reviews.MethodsWith the approval of NHS England, we conducted a cohort study of 20 million adult patient records in general practice, in‐situ using the OpenSAFELY platform. For each month, between April 2019 and March 2022, we report the percentage of patients with a medication review coded monthly and in the previous 12 months with breakdowns by regional, clinical and demographic subgroups and those prescribed high‐risk medications.ResultsIn April 2019, 32.3% of patients had a medication review coded in the previous 12 months. During the first COVID‐19 lockdown, monthly activity decreased (−21.1% April 2020), but the 12‐month rate was not substantially impacted (−10.5% March 2021). The rate of structured medication review in the last 12 months reached 2.9% by March 2022, with higher percentages in high‐risk groups (care home residents 34.1%, age 90+ years 13.1%, high‐risk medications 10.2%). The most used medication review code was Medication review done 314530002 (59.5%).ConclusionsThere was a substantial reduction in the monthly rate of medication reviews during the pandemic but rates recovered by the end of the study period. Structured medication reviews were prioritized for high‐risk patients.

We present Emu Video, a text-to-video generation model that factorizes the generation into two steps: first generating an image conditioned on the text, and then generating a video conditioned on the text and the generated image. We identify critical design decisions--adjusted noise schedules for diffusion, and multi-stage training--that enable us to directly generate high quality and high resolution videos, without requiring a deep cascade of models as in prior work. In human evaluations, our generated videos are strongly preferred in quality compared to all prior work--81% vs. Google's Imagen Video, 90% vs. Nvidia's PYOCO, and 96% vs. Meta's Make-A-Video. Our model outperforms commercial solutions such as RunwayML's Gen2 and Pika Labs. Finally, our factorizing approach naturally lends itself to animating images based on a user's text prompt, where our generations are preferred 96% over prior work.

ObjectiveTo investigate the effect of the covid-19 pandemic on the number of patients with group A streptococcal infections and related antibiotic prescriptions.DesignRetrospective cohort study in England using OpenSAFELY-TPP.SettingPrimary care practices in England that used TPP SystmOne software, 1 January 2018 to 31 March 2023, with the approval of NHS England.ParticipantsPatients registered at a TPP practice at the start of each month of the study period. Patients with missing data for sex or age were excluded, resulting in a population of 23 816 470 in January 2018, increasing to 25 541 940 by March 2023.Main outcome measuresMonthly counts and crude rates of patients with group A streptococcal infections (sore throat or tonsillitis, scarlet fever, and invasive group A streptococcal infections), and recommended firstline, alternative, and reserved antibiotic prescriptions linked with a group A streptococcal infection before (pre-April 2020), during, and after (post-April 2021) covid-19 restrictions. Maximum and minimum count and rate for each infectious season (time from September to August), as well as the rate ratio of the 2022-23 season compared with the last comparably high season (2017-18).ResultsThe number of patients with group A streptococcal infections, and antibiotic prescriptions linked to an indication of group A streptococcal infection, peaked in December 2022, higher than the peak in 2017-18. The rate ratios for monthly sore throat or tonsillitis (possible group A streptococcal throat infection), scarlet fever, and invasive group A streptococcal infection in 2022-23 relative to 2017-18 were 1.39 (95% confidence interval (CI) 1.38 to 1.40), 2.68 (2.59 to 2.77), and 4.37 (2.94 to 6.48), respectively. The rate ratio for prescriptions of first line, alternative, and reserved antibiotics to patients with group A streptococcal infections in 2022-23 relative to 2017-18 were 1.37 (95% CI 1.35 to 1.38), 2.30 (2.26 to 2.34), and 2.42 (2.24 to 2.61), respectively. For individual antibiotic prescriptions in 2022-23, azithromycin showed the greatest relative increase versus 2017-18, with a rate ratio of 7.37 (6.22 to 8.74). This finding followed a marked decrease in the recording of patients with group A streptococcal infections and associated prescriptions during the period of covid-19 restrictions where the maximum count and rates were lower than any minimum rates before the covid-19 pandemic.ConclusionsRecording of rates of scarlet fever, sore throat or tonsillitis, and invasive group A streptococcal infections, and associated antibiotic prescribing, peaked in December 2022. Primary care data can supplement existing infectious disease surveillance through linkages with relevant prescribing data and detailed analysis of clinical and demographic subgroups.

The VoxCeleb Speaker Recognition Challenges (VoxSRC) were a series of challenges and workshops that ran annually from 2019 to 2023. The challenges primarily evaluated the tasks of speaker recognition and diarisation under various settings including: closed and open training data; as well as supervised, self-supervised, and semi-supervised training for domain adaptation. The challenges also provided publicly available training and evaluation datasets for each task and setting, with new test sets released each year. In this paper, we provide a review of these challenges that covers: what they explored; the methods developed by the challenge participants and how these evolved; and also the current state of the field for speaker verification and diarisation. We chart the progress in performance over the five installments of the challenge on a common evaluation dataset and provide a detailed analysis of how each year's special focus affected participants' performance. This paper is aimed both at researchers who want an overview of the speaker recognition and diarisation field, and also at challenge organisers who want to benefit from the successes and avoid the mistakes of the VoxSRC challenges. We end with a discussion of the current strengths of the field and open challenges.

We present Movie Gen, a cast of foundation models that generates high-quality, 1080p HD videos with different aspect ratios and synchronized audio. We also show additional capabilities such as precise instruction-based video editing and generation of personalized videos based on a user's image. Our models set a new state-of-the-art on multiple tasks: text-to-video synthesis, video personalization, video editing, video-to-audio generation, and text-to-audio generation. Our largest video generation model is a 30B parameter transformer trained with a maximum context length of 73K video tokens, corresponding to a generated video of 16 seconds at 16 frames-per-second. We show multiple technical innovations and simplifications on the architecture, latent spaces, training objectives and recipes, data curation, evaluation protocols, parallelization techniques, and inference optimizations that allow us to reap the benefits of scaling pre-training data, model size, and training compute for training large scale media generation models. We hope this paper helps the research community to accelerate progress and innovation in media generation models. All videos from this paper are available at https://go.fb.me/MovieGenResearchVideos.

BackgroundThe COVID-19 pandemic disrupted health-care delivery, including difficulty accessing in-person care, which could have increased the need for strong pharmacological pain relief. Due to the risks associated with overprescribing of opioids, especially to vulnerable populations, we aimed to quantify changes to measures during the COVID-19 pandemic, overall, and by key subgroups.MethodsFor this interrupted time-series analysis study conducted in England, with National Health Service England approval, we used routine clinical data from more than 20 million general practice adult patients in OpenSAFELY-TPP, which is a a secure software platform for analysis of electronic health records. We included all adults registered with a primary care practice using TPP-SystmOne software. Using interrupted time-series analysis, we quantified prevalent and new opioid prescribing before the COVID-19 pandemic (January, 2018–February, 2020), during the lockdown (March, 2020–March, 2021), and recovery periods (April, 2021–June, 2022), overall and stratified by demographics (age, sex, deprivation, ethnicity, and geographical region) and in people in care homes identified via an address-matching algorithm.FindingsThere was little change in prevalent prescribing during the pandemic, except for a temporary increase in March, 2020. We observed a 9·8% (95% CI –14·5 to –6·5) reduction in new opioid prescribing from March, 2020, with a levelling of the downward trend, and rebounding slightly after April, 2021 (4·1%, 95% CI –0·9 to 9·4). Opioid prescribing rates varied by demographics, but we found a reduction in new prescribing for all subgroups except people aged 80 years or older. Among care home residents, in April, 2020, parenteral opioid prescribing increased by 186·3% (153·1 to 223·9).InterpretationOpioid prescribing increased temporarily among older people and care home residents, likely reflecting use to treat end-of-life COVID-19 symptoms. Despite vulnerable populations being more affected by health-care disruptions, disparities in opioid prescribing by most demographic subgroups did not widen during the pandemic. Further research is needed to understand what is driving the changes in new opioid prescribing and its relation to changes to health-care provision during the pandemic.FundingThe Wellcome Trust, Medical Research Council, The National Institute for Health and Care Research, UK Research and Innovation, and Health Data Research UK.

Background: The COVID-19 pandemic led to disruptions in healthcare delivery, including postponement of elective procedures and difficulty accessing in-person care, which may have increased the need for strong pharmacological pain relief in some patients. Methods: With NHS England approval, we used routine clinical data from >20 million general practice adult patients in OpenSAFELY-TPP. We used interrupted time series analysis to quantify trends in prevalent and incident opioid prescribing prior to the pandemic (January 2018-February 2020) and changes during the COVID-19 lockdown period (March 2020-March 2021) and recovery period (April 2021-June 2022). We identified how these changes varied in people living in care homes, and by age, sex, deprivation, ethnicity, and geographic region. Results: The median number of people prescribed an opioid per month was 50.9 per 1000 patients prior to the pandemic. We observed little change in overall prescribing after the start of the pandemic, except for a temporary increase in March 2020. There was a 9.8% (95%CI -14.5%, -6.5%) reduction in new opioid prescribing from March 2020, sustained to the end of the study period. Reductions in new prescribing were observed for all demographics except people 80+ years. Among care home residents, in April 2020 new opioid prescribing increased by 112.5% (95%CI 92.2%, 134.9%) and parenteral opioid prescribing increased by 186.3% (95%CI 153.1%, 223.9%). Conclusion: Changes in opioid prescribing during the COVID-19 pandemic were mostly consistent across subgroups with the exception of differences by age and care home residence. Among people in care homes, increases in parenteral opioid prescribing likely reflect use to treat end-of-life COVID-19 symptoms. Further research is needed to understand what is driving the reduction in new opioid prescribing and its relation to changes to health care provision during the pandemic. ### Competing Interest Statement BG has received research funding from the Bennett Foundation, the Laura and John Arnold Foundation, the NHS National Institute for Health Research (NIHR), the NIHR School of Primary Care Research, NHS England, the NIHR Oxford Biomedical Research Centre, the Mohn-Westlake Foundation, NIHR Applied Research Collaboration Oxford and Thames Valley, the Wellcome Trust, the Good Thinking Foundation, Health Data Research UK, the Health Foundation, the World Health Organisation, UKRI MRC, Asthma UK, the British Lung Foundation, and the Longitudinal Health and Wellbeing strand of the National Core Studies programme; he has previously been a Non-Executive Director at NHS Digital; he also receives personal income from speaking and writing for lay audiences on the misuse of science. BMK is also employed by NHS England working on medicines policy and clinical lead for primary care medicines data. AM has represented the RCGP in the health informatics group and the Profession Advisory Group that advises on access to GP Data for Pandemic Planning and Research (GDPPR); the latter was a paid role. AM is a former employee and interim Chief Medical Officer of NHS Digital. AM has consulted for health care vendors, the last time in 2022; the companies consulted in the last 3 years have no relationship to OpenSAFELY. ### Funding Statement The OpenSAFELY Platform is supported by grants from the Wellcome Trust (222097/Z/20/Z) and MRC (MR/V015737/1, MC\_PC\_20059, MR/W016729/1). In addition, development of OpenSAFELY has been funded by the Longitudinal Health and Wellbeing strand of the National Core Studies programme (MC\_PC\_20030: MC\_PC\_20059), the NIHR funded CONVALESCENCE programme (COV-LT-0009), NIHR (NIHR135559, COV-LT2-0073), and the Data and Connectivity National Core Study funded by UK Research and Innovation (MC\_PC\_20058) and Health Data Research UK ### Author Declarations I confirm all relevant ethical guidelines have been followed, and any necessary IRB and/or ethics committee approvals have been obtained. Yes The details of the IRB/oversight body that provided approval or exemption for the research described are given below: This study was approved by the Health Research Authority (Research Ethics Committee reference 20/LO/0651) and by the London School of Hygiene and Tropical Medicine Ethics Board (reference 21863). I confirm that all necessary patient/participant consent has been obtained and the appropriate institutional forms have been archived, and that any patient/participant/sample identifiers included were not known to anyone (e.g., hospital staff, patients or participants themselves) outside the research group so cannot be used to identify individuals. Yes I understand that all clinical trials and any other prospective interventional studies must be registered with an ICMJE-approved registry, such as ClinicalTrials.gov. I confirm that any such study reported in the manuscript has been registered and the trial registration ID is provided (note: if posting a prospective study registered retrospectively, please provide a statement in the trial ID field explaining why the study was not registered in advance). Yes I have followed all appropriate research reporting guidelines, such as any relevant EQUATOR Network research reporting checklist(s) and other pertinent material, if applicable. Yes Primary care records managed by the GP software provider, TPP were linked to Office of National Statistics (ONS) death data through OpenSAFELY and were linked, stored and analysed securely within the OpenSAFELY platform: https://opensafely.org/ as part of the NHS England OpenSAFELY COVID-19 service. Data include pseudonymised data such as coded diagnoses, medications and physiological parameters. No free text data are included. All code is shared openly for review and re-use under MIT open license (https://github.com/opensafely/opioids-covid-research). Detailed pseudonymised patient data are potentially re-identifiable and therefore not shared.

Aims The COVID‐19 pandemic created unprecedented pressure on healthcare services. This study investigates whether disease‐modifying antirheumatic drug (DMARD) safety monitoring was affected during the COVID‐19 pandemic. Methods A population‐based cohort study was conducted using the OpenSAFELY platform to access electronic health record data from 24.2 million patients registered at general practices using TPP's SystmOne software. Patients were included for further analysis if prescribed azathioprine, leflunomide or methotrexate between November 2019 and July 2022. Outcomes were assessed as monthly trends and variation between various sociodemographic and clinical groups for adherence with standard safety monitoring recommendations. Results An acute increase in the rate of missed monitoring occurred across the study population (+12.4 percentage points) when lockdown measures were implemented in March 2020. This increase was more pronounced for some patient groups (70–79 year‐olds: +13.7 percentage points; females: +12.8 percentage points), regions (North West: +17.0 percentage points), medications (leflunomide: +20.7 percentage points) and monitoring tests (blood pressure: +24.5 percentage points). Missed monitoring rates decreased substantially for all groups by July 2022. Consistent differences were observed in overall missed monitoring rates between several groups throughout the study. Conclusion DMARD monitoring rates temporarily deteriorated during the COVID‐19 pandemic. Deterioration coincided with the onset of lockdown measures, with monitoring rates recovering rapidly as lockdown measures were eased. Differences observed in monitoring rates between medications, tests, regions and patient groups highlight opportunities to tackle potential inequalities in the provision or uptake of monitoring services. Further research should evaluate the causes of the differences identified between groups.

Background The COVID-19 pandemic disrupted cardiovascular disease management in primary care in England.Objective To describe the impact of the pandemic on blood pressure screening and hypertension management based on a national quality of care scheme (Quality and Outcomes Framework, QOF) across key demographic, regional and clinical subgroups.Methods With NHS England approval, a population-based cohort study was conducted using OpenSAFELY-TPP on 25.2 million NHS patients registered at general practices (March 2019 to March 2023). We examined monthly changes in recorded blood pressure screening in the preceding 5 years in patients aged >= 45 years and recorded the hypertension prevalence and the percentage of patients treated to target (<= 140/90 mmHg for patients aged <= 79 years and <= 150/90 mmHg for patients aged >= 80 years) in the preceding 12 months.Results The percentage of patients aged >= 45 years who had blood pressure screening recorded in the preceding 5 years decreased from 90% (March 2019) to 85% (March 2023). Recorded hypertension prevalence was relatively stable at 15% throughout the study period. The percentage of patients with a record of hypertension treated to target in the preceding 12 months reduced from a maximum of 71% (March 2020) to a minimum of 47% (February 2021) in patients aged <= 79 years and from 85% (March 2020) to a minimum of 58% (February 2021) in patients aged >= 80 years before recovery. Blood pressure screening rates in the preceding 5 years remained stable in older people, patients with recorded learning disability or care home status.Conclusions The pandemic substantially disrupted hypertension management QOF indicators, which is likely attributable to general reductions of blood pressure measurement including screening. OpenSAFELY can be used to continuously monitor changes in national quality-of-care schemes to identify changes in key clinical subgroups early and support prioritisation of recovery from care disrupted by COVID-19.

Background: The coronavirus disease 2019 (COVID-19) vaccination programme in England was extended to include all adolescents and children by April 2022. The aim of this paper is to describe trends and variation in vaccine coverage in different clinical and demographic groups amongst adolescents and children in England by August 2022. Methods: With the approval of NHS England, a cohort study was conducted of 3.21 million children and adolescents’ records in general practice in England, in situ and within the infrastructure of the electronic health record software vendor TPP using OpenSAFELY. Vaccine coverage across various demographic (sex, deprivation index and ethnicity) and clinical (risk status) populations is described. Results: Coverage is higher amongst adolescents than it is amongst children, with 53.5% adolescents and 10.8% children having received their first dose of the COVID-19 vaccine. Within those groups, coverage varies by ethnicity, deprivation index and risk status; there is no evidence of variation by sex. Conclusion: First dose COVID-19 vaccine coverage is shown to vary amongst various demographic and clinical groups of children and adolescents.

Background: The coronavirus disease 2019 (COVID-19) vaccination programme in England was extended to include all adolescents and children by April 2022. The aim of this paper is to describe trends and variation in vaccine coverage in different clinical and demographic groups amongst adolescents and children in England by August 2022. Methods: With the approval of NHS England, a cohort study was conducted of 3.21 million children and adolescents’ records in general practice in England, in situ and within the infrastructure of the electronic health record software vendor TPP using OpenSAFELY. Vaccine coverage across various demographic (sex, deprivation index and ethnicity) and clinical (risk status) populations is described. Results: Coverage is higher amongst adolescents than it is amongst children, with 53.5% adolescents and 10.8% children having received their first dose of the COVID-19 vaccine. Within those groups, coverage varies by ethnicity, deprivation index and risk status; there is no evidence of variation by sex. Conclusion: First dose COVID-19 vaccine coverage is shown to vary amongst various demographic and clinical groups of children and adolescents.

Speaker embedding extractors (EEs), which map input audio to a speaker discriminant latent space, are of paramount importance in speaker diarisation. However, there are several challenges when adopting EEs for diarisation, from which we tackle two key problems. First, the evaluation is not straightforward because the required features differ between speaker verification and diarisation. We show that better performance on widely adopted speaker verification evaluation protocols does not lead to better diarisation performance. Second, embedding extractors have not seen utterances in which multiple speakers exist. These inputs are inevitably present in speaker diarisation because of overlapped speech and speaker changes; they degrade the performance. To mitigate the first problem, we generate speaker verification evaluation protocols that better mimic the diarisation scenario. We propose two data augmentation techniques to alleviate the second problem, making embedding extractors aware of overlapped speech or speaker change input. One technique generates overlapped speech segments, and the other generates segments where two speakers utter sequentially. Extensive experimental results using three state-of-the-art speaker embedding extractors demonstrate that both proposed approaches are effective.

Background: Approaches to addressing unwarranted variation in health care service delivery have traditionally relied on the prospective identification of activities and outcomes, based on a hypothesis, with subsequent reporting against defined measures. Practice-level prescribing data in England are made publicly available by the National Health Service (NHS) Business Services Authority for all general practices. There is an opportunity to adopt a more data-driven approach to capture variability and identify outliers by applying hypothesis-free, data-driven algorithms to national data sets. Objective: This study aimed to develop and apply a hypothesis-free algorithm to identify unusual prescribing behavior in primary care data at multiple administrative levels in the NHS in England and to visualize these results using organization-specific interactive dashboards, thereby demonstrating proof of concept for prioritization approaches. Methods: Here we report a new data-driven approach to quantify how "unusual" the prescribing rates of a particular chemical within an organization are as compared to peer organizations, over a period of 6 months (June-December 2021). This is followed by a ranking to identify which chemicals are the most notable outliers in each organization. These outlying chemicals are calculated for all practices, primary care networks, clinical commissioning groups, and sustainability and transformation partnerships in England. Our results are presented via organization-specific interactive dashboards, the iterative development of which has been informed by user feedback. Results: We developed interactive dashboards for every practice (n=6476) in England, highlighting the unusual prescribing of 2369 chemicals (dashboards are also provided for 42 sustainability and transformation partnerships, 106 clinical commissioning groups, and 1257 primary care networks). User feedback and internal review of case studies demonstrate that our methodology identifies prescribing behavior that sometimes warrants further investigation or is a known issue. Conclusions: Data-driven approaches have the potential to overcome existing biases with regard to the planning and execution of audits, interventions, and policy making within NHS organizations, potentially revealing new targets for improved health care service delivery. We present our dashboards as a proof of concept for generating candidate lists to aid expert users in their interpretation of prescribing data and prioritize further investigations and qualitative research in terms of potential targets for improved performance.

We introduce MoCA, a Motion-Conditioned Image Animation approach for video editing. It leverages a simple decomposition of the video editing problem into image editing followed by motion-conditioned image animation. Furthermore, given the lack of robust evaluation datasets for video editing, we introduce a new benchmark that measures edit capability across a wide variety of tasks, such as object replacement, background changes, style changes, and motion edits. We present a comprehensive human evaluation of the latest video editing methods along with MoCA, on our proposed benchmark. MoCA establishes a new state-of-the-art, demonstrating greater human preference win-rate, and outperforming notable recent approaches including Dreamix (63%), MasaCtrl (75%), and Tune-A-Video (72%), with especially significant improvements for motion edits.

This paper summarises the findings from the VoxCeleb Speaker Recognition Challenge 2022 (VoxSRC-22), which was held in conjunction with INTERSPEECH 2022. The goal of this challenge was to evaluate how well state-of-the-art speaker recognition systems can diarise and recognise speakers from speech obtained "in the wild". The challenge consisted of: (i) the provision of publicly available speaker recognition and diarisation data from YouTube videos together with ground truth annotation and standardised evaluation software; and (ii) a public challenge and hybrid workshop held at INTERSPEECH 2022. We describe the four tracks of our challenge along with the baselines, methods, and results. We conclude with a discussion on the new domain-transfer focus of VoxSRC-22, and on the progression of the challenge from the previous three editions.

The third instalment of the VoxCeleb Speaker Recognition Challenge was held in conjunction with Interspeech 2021. The aim of this challenge was to assess how well current speaker recognition technology is able to diarise and recognise speakers in unconstrained or `in the wild' data. The challenge consisted of: (i) the provision of publicly available speaker recognition and diarisation data from YouTube videos together with ground truth annotation and standardised evaluation software; and (ii) a virtual public challenge and workshop held at Interspeech 2021. This paper outlines the challenge, and describes the baselines, methods and results. We conclude with a discussion on the new multi-lingual focus of VoxSRC 2021, and on the progression of the challenge since the previous two editions.