MutualNet: Adaptive ConvNet via Mutual Learning From Different Model Configurations.

Most existing deep neural networks are static, which means they can only perform inference at a fixed complexity. But the resource budget can vary substantially across different devices. Even on a single device, the affordable budget can change with different scenarios, and repeatedly training networks for each required budget would be incredibly expensive. Therefore, in this work, we propose a general method called MutualNet to train a single network that can run at a diverse set of resource constraints. Our method trains a cohort of model configurations with various network widths and input resolutions. This mutual learning scheme not only allows the model to run at different width-resolution configurations but also transfers the unique knowledge among these configurations, helping the model to learn stronger representations overall. MutualNet is a general training methodology that can be applied to various network structures (e.g., 2D networks: MobileNets, ResNet, 3D networks: SlowFast, X3D) and various tasks (e.g., image classification, object detection, segmentation, and action recognition), and is demonstrated to achieve consistent improvements on a variety of datasets. Since we only train the model once, it also greatly reduces the training cost compared to independently training several models. Surprisingly, MutualNet can also be used to significantly boost the performance of a single network, if dynamic resource constraints are not a concern. In summary, MutualNet is a unified method for both static and adaptive, 2D and 3D networks. Code and pre-trained models are available at https://github.com/taoyang1122/MutualNet.

COVID-19 within a large UK prison with a high number of vulnerable adults, march to june 2020: An outbreak investigation and screening event.

OBJECTIVE: To describe the public health response to COVID-19 within a large prison with a high number of clinically vulnerable residents. DESIGN: An outbreak investigation was undertaken among all residents and staff. A screening event involved nose and throat swab samples from residents and staff, examined by polymerase chain reaction (PCR). An electronic questionnaire regarding risk factors was distributed to staff. RESULTS: 58 residents out of 1,156 (5.0%, 95% CI: 3.8 - 6.3) and 129 staff out of 510 (25.3%, 95% CI: 21.5 - 29.1) displayed COVID-19 symptoms, including six and eight confirmed, respectively. Residents reported cough symptoms with no fever (29.3%), followed by a cough and fever (15.5%). 62.1% of symptomatic residents were 50 years or older, placing them in the group at risk for severe COVID-19 disease. Wing I had the highest attack rate (12.5%). 1,063 individuals were swabbed during the 5-day screening event, and all had negative swab results. CONCLUSION: The findings were consistent with the hypothesis of a propagating outbreak with decreasing incidence since the peak date of onset. COVID-19 transmission within a high-risk setting was quickly contained, and an explosive outbreak was prevented through a multi-agency public health response.

The evaluation of a standardized call/recall system for childhood immunizations in Wandsworth, England.

To improve uptake of childhood immunizations in Wandsworth we developed a standardized call/recall system based on parents being sent three reminders and defaulters being referred to a Health Visitor. Thirty-two out of 44 primary care practices in the area implemented the intervention in September 2011. The aim of this study was to evaluate the implementation, delivery and impact on immunization uptake of the new call/recall system. To assess implementation and delivery, a mixed method approach was used including qualitative (structured interviews) and quantitative (data collected at three months post-implementation) assessment. To assess the impact, we used Student's t test to compare the difference in immunization uptake rates between intervention and non-intervention practices before and after implementation. The call/recall system was viewed positively by both parents and staff. Most children due or overdue immunizations were successfully captured by the 1st invitation reminder. After three invitations, between 87.3 % (MMR1) and 92.2 % (pre-school booster) of children identified as due or overdue immunizations successfully responded. Prior to implementation there was no difference in uptake rates between intervention and non-intervention practices. Post-implementation uptake rates for DTaP/IPV/Hib, MMR1, MMR2 and the pre-school booster were significantly greater in the intervention practices. Similar findings were seen for PCV and Hib/MenC boosters, although the differences were not statistically significant at the 5 % level. Following the successful implementation of a standardized call/recall system in Wandsworth, other regions or primary care practices may wish to consider introducing a similar system to help improve their immunization coverage levels.

Emergence of a new antibiotic resistance mechanism in India, Pakistan, and the UK: a molecular, biological, and epidemiological study.

BACKGROUND: Gram-negative Enterobacteriaceae with resistance to carbapenem conferred by New Delhi metallo-beta-lactamase 1 (NDM-1) are potentially a major global health problem. We investigated the prevalence of NDM-1, in multidrug-resistant Enterobacteriaceae in India, Pakistan, and the UK. METHODS: Enterobacteriaceae isolates were studied from two major centres in India--Chennai (south India), Haryana (north India)--and those referred to the UK's national reference laboratory. Antibiotic susceptibilities were assessed, and the presence of the carbapenem resistance gene bla(NDM-1) was established by PCR. Isolates were typed by pulsed-field gel electrophoresis of XbaI-restricted genomic DNA. Plasmids were analysed by S1 nuclease digestion and PCR typing. Case data for UK patients were reviewed for evidence of travel and recent admission to hospitals in India or Pakistan. FINDINGS: We identified 44 isolates with NDM-1 in Chennai, 26 in Haryana, 37 in the UK, and 73 in other sites in India and Pakistan. NDM-1 was mostly found among Escherichia coli (36) and Klebsiella pneumoniae (111), which were highly resistant to all antibiotics except to tigecycline and colistin. K pneumoniae isolates from Haryana were clonal but NDM-1 producers from the UK and Chennai were clonally diverse. Most isolates carried the NDM-1 gene on plasmids: those from UK and Chennai were readily transferable whereas those from Haryana were not conjugative. Many of the UK NDM-1 positive patients had travelled to India or Pakistan within the past year, or had links with these countries. INTERPRETATION: The potential of NDM-1 to be a worldwide public health problem is great, and co-ordinated international surveillance is needed.

The burden of alcohol-related ill health in the United Kingdom.

BACKGROUND: Although moderate alcohol consumption has been shown to confer a protective effect for specific diseases, current societal patterns of alcohol use impose a huge health and economic burden on modern society. This study presents a method for estimating the health and economic burden of alcohol consumption to the UK National Health Service (NHS). METHODS: Previous estimates of NHS costs attributable to alcohol consumption were identified by systematic literature review. The mortality and morbidity due to alcohol consumption was calculated using information from the World Health Organization Global Burden of Disease Project and routinely collected mortality data. Direct health-care costs were derived using information on population attributable fractions for conditions related to alcohol consumption and NHS cost data. RESULTS: We estimate that alcohol consumption was responsible for 31,000 deaths in the UK in 2005 and that alcohol consumption cost the UK NHS 3.0 billion pounds in 2005-06. Alcohol consumption was responsible for 10% of all disability adjusted life years in 2002 (male: 15%; female: 4%) in the UK. CONCLUSIONS: Alcohol consumption is a considerable public health burden in the UK. The comparison of the health and economic burden of various lifestyle factors is essential in prioritizing and resourcing public health action.

Trends in overweight and obesity among 5-7-year-old White and South Asian children born between 1991 and 1999.

OBJECTIVE: To determine the trends in overweight and obesity among White and South Asian children aged 5-7 years born between 1991 and 1999 and included in the East Berkshire Child Health System. METHODS: Children were grouped into nine cohorts based on their year of birth. The UK National BMI percentile classification was used to classify the children as overweight and obese and to examine the prevalence and trends by year of birth, sex and ethnicity. RESULTS: Overall, more boys (10.1%; 9.7-10.6%) than girls (9.1%; 8.7-9.6%) were obese (P < 0.003). South Asian boys were more likely to be overweight (OR 1.92; 95% CI 1.62-2.28; P < 0.01) and obese (OR 1.53; 95% CI 1.28-1.89; P < 0.01) than South Asian girls. Overweight (1.77; 1.56-2.00; P < 0.05) and obesity (1.76; 1.50-2.06; P < 0.05) were significantly higher among South Asian boys compared with their White counterparts (baseline). After adjusting for sex, ethnicity and year of birth, South Asian children were 27% more overweight (P < 0.01) and 45% more obese (P < 0.01) compared with White children, and boys were 6% more overweight (P = 0.04) and 12% more obese (P = 0.003) compared with girls. There was an increasing trend in overweight among boys (P = 0.01) and girls (P = 0.003); and in obesity among boys (P < 0.001) and girls (P = 0.008) in children born from 1991 to 1999. CONCLUSION: There is a significant rise in childhood obesity among 5-7-year-old children. Overweight and obesity among South Asian boys are significantly higher than that among South Asian girls. This group may be at greater risk of morbidity and mortality related to obesity and may need to be targeted appropriately for interventions to reduce obesity.