Sepsis is the third leading cause of death worldwide and the main cause of mortality in hospitals
, but the best treatment strategy remains uncertain. In particular, evidence suggests that current ...practices in the administration of intravenous fluids and vasopressors are suboptimal and likely induce harm in a proportion of patients
. To tackle this sequential decision-making problem, we developed a reinforcement learning agent, the Artificial Intelligence (AI) Clinician, which extracted implicit knowledge from an amount of patient data that exceeds by many-fold the life-time experience of human clinicians and learned optimal treatment by analyzing a myriad of (mostly suboptimal) treatment decisions. We demonstrate that the value of the AI Clinician's selected treatment is on average reliably higher than human clinicians. In a large validation cohort independent of the training data, mortality was lowest in patients for whom clinicians' actual doses matched the AI decisions. Our model provides individualized and clinically interpretable treatment decisions for sepsis that could improve patient outcomes.
To systematically examine the design, reporting standards, risk of bias, and claims of studies comparing the performance of diagnostic deep learning algorithms for medical imaging with that of expert ...clinicians.
Systematic review.
Medline, Embase, Cochrane Central Register of Controlled Trials, and the World Health Organization trial registry from 2010 to June 2019.
Randomised trial registrations and non-randomised studies comparing the performance of a deep learning algorithm in medical imaging with a contemporary group of one or more expert clinicians. Medical imaging has seen a growing interest in deep learning research. The main distinguishing feature of convolutional neural networks (CNNs) in deep learning is that when CNNs are fed with raw data, they develop their own representations needed for pattern recognition. The algorithm learns for itself the features of an image that are important for classification rather than being told by humans which features to use. The selected studies aimed to use medical imaging for predicting absolute risk of existing disease or classification into diagnostic groups (eg, disease or non-disease). For example, raw chest radiographs tagged with a label such as pneumothorax or no pneumothorax and the CNN learning which pixel patterns suggest pneumothorax.
Adherence to reporting standards was assessed by using CONSORT (consolidated standards of reporting trials) for randomised studies and TRIPOD (transparent reporting of a multivariable prediction model for individual prognosis or diagnosis) for non-randomised studies. Risk of bias was assessed by using the Cochrane risk of bias tool for randomised studies and PROBAST (prediction model risk of bias assessment tool) for non-randomised studies.
Only 10 records were found for deep learning randomised clinical trials, two of which have been published (with low risk of bias, except for lack of blinding, and high adherence to reporting standards) and eight are ongoing. Of 81 non-randomised clinical trials identified, only nine were prospective and just six were tested in a real world clinical setting. The median number of experts in the comparator group was only four (interquartile range 2-9). Full access to all datasets and code was severely limited (unavailable in 95% and 93% of studies, respectively). The overall risk of bias was high in 58 of 81 studies and adherence to reporting standards was suboptimal (<50% adherence for 12 of 29 TRIPOD items). 61 of 81 studies stated in their abstract that performance of artificial intelligence was at least comparable to (or better than) that of clinicians. Only 31 of 81 studies (38%) stated that further prospective studies or trials were required.
Few prospective deep learning studies and randomised trials exist in medical imaging. Most non-randomised trials are not prospective, are at high risk of bias, and deviate from existing reporting standards. Data and code availability are lacking in most studies, and human comparator groups are often small. Future studies should diminish risk of bias, enhance real world clinical relevance, improve reporting and transparency, and appropriately temper conclusions.
PROSPERO CRD42019123605.
Air travel during the COVID-19 pandemic is challenging for travellers, airlines, airports, health authorities, and governments. We reviewed multiple aspects of COVID peri-pandemic air travel, ...including data on traveller numbers, peri-flight prevention, and testing recommendations and in-flight SARS-CoV-2 transmission, photo-epidemiology of mask use, the pausing of air travel to mass gathering events, and quarantine measures and their effectiveness.
Flights are reduced by 43% compared to 2019. Hygiene measures, mask use, and distancing are effective, while temperature screening has been shown to be unreliable. Although the risk of in-flight transmission is considered to be very low, estimated at one case per 27 million travellers, confirmed in-flight cases have been published. Some models exist and predict minimal risk but fail to consider human behavior and airline procedures variations. Despite aircraft high-efficiency filtering, there is some evidence that passengers within two rows of an index case are at higher risk. Air travel to mass gatherings should be avoided. Antigen testing is useful but impaired by time lag to results. Widespread application of solutions such as saliva-based, rapid testing or even detection with the help of “sniffer dogs” might be the way forward. The “traffic light system” for traveling, recently introduced by the Council of the European Union is a first step towards normalization of air travel. Quarantine of travellers may delay introduction or re-introduction of the virus, or may delay the peak of transmission, but the effect is small and there is limited evidence. New protocols detailing on-arrival, rapid testing and tracing are indicated to ensure that restricted movement is pragmatically implemented. Guidelines from airlines are non-transparent. Most airlines disinfect their flights and enforce wearing masks and social distancing to a certain degree. A layered approach of non-pharmaceutical interventions, screening and testing procedures, implementation and adherence to distancing, hygiene measures and mask use at airports, in-flight and throughout the entire journey together with pragmatic post-flight testing and tracing are all effective measures that can be implemented.
Ongoing research and systematic review are indicated to provide evidence on the utility of preventive measures and to help answer the question “is it safe to fly?“.
Pneumonia is responsible for approximately 230,000 deaths in Europe, annually. Comprehensive and comparable reports on pneumonia mortality trends across the European Union (EU) are lacking.
A ...temporal analysis of national mortality statistics to compare trends in pneumonia age-standardised death rates (ASDR) of EU countries between 2001 and 2014 was performed. International Classification of Diseases version 10 (ICD-10) codes were used to extract data from the World Health Organisation European Detailed Mortality Database and trends were analysed using Joinpoint regression.
Median pneumonia mortality across the EU for the last recorded observation was 19.8 / 100,000 and 6.9 / 100,000 for males and females, respectively. Mortality was higher in males across all EU countries, most notably in Estonia and Lithuania where the ratio of male to female ASDR was 4.0 and 3.7, respectively. Gender mortality differences were lowest in the UK and Demark with ASDR ratios of 1.1 and 1.5, respectively. Pneumonia mortality across all countries decreased by a median of 31.0% over the observation period. Countries that demonstrated an increase in pneumonia mortality were Poland (males + 33.1%, females + 10.2%), and Lithuania (males + 6.0%).
Mortality from pneumonia is improving in most EU countries, however substantial variation in trends remains between countries and between genders.
Celotno besedilo
Dostopno za:
DOBA, IZUM, KILJ, NUK, PILJ, PNG, SAZU, SIK, UILJ, UKNU, UL, UM, UPUK
Simulation-based approaches to disease progression allow us to make counterfactual predictions about the effects of an untried series of treatment choices. However, building accurate simulators of ...disease progression is challenging, limiting the utility of these approaches for real world treatment planning. In this work, we present a novel simulation-based reinforcement learning approach that mixes between models and kernel-based approaches to make its forward predictions. On two real world tasks, managing sepsis and treating HIV, we demonstrate that our approach both learns state-of-the-art treatment policies and can make accurate forward predictions about the effects of treatments on unseen patients.
Celotno besedilo
Dostopno za:
DOBA, IZUM, KILJ, NUK, PILJ, PNG, SAZU, SIK, UILJ, UKNU, UL, UM, UPUK
Introduction: Since Gagarin became the first human to travel into space and complete one orbit around the Earth, on 12 April 1961, the number of manned spaceflights has increased significantly. ...Spaceflight is still complex and has potential risk for incidents and accidents. The aim of this study was to analyze how safe it is for humans to travel in space. Objectives: This paper, therefore, summarizes incidents and accidents covering the six decades of manned spaceflight (1961–2020). Material and methods: Extensive PubMed, Cochrane, and Google Scholar searches were made with search strings of “incidents”, “accident”, “spaceflight”, and “orbit”, and including all vehicles so far. Search terms were combined by AND or OR in search strings. Of the results obtained, studies which evaluated manned spaceflight were included in the study. Data from the National Aeronautics Space Agency (NASA), the Russian Space Agency (ROSCOSMOS), the European Space Agency (ESA), and the Chinese Space Agency (CNSA), as well as from the Virgin Galactic and the SpaceX databases, were searched to complete data and to identify all the accomplished manned spaceflights, as well as all incidents and accidents that have occurred in the specific period. Search results were compared to findings on Wikipedia, Encyclopedia Astronautica, and other public webpages. Reference lists of included articles/homepages were also included for further potential data. Results: From 1961–2020, our data revealed an increasing number of manned space flights, n = 327. The number of times an astronaut has been sent to space, n = 1294, resulted in an accumulated n = 19,414 days spent in space. The number of days spent in orbit has constantly increased from 1961 until today. The number of incidents (altogether n = 36) and accidents (altogether n = 5) has constantly decreased. The number of astronauts who have died during spaceflight is represented by n = 19. The current statistical fatality rate is 5.8% (deaths per spaceflight) with the highest fatality rate in the 1960s (0.013 deaths/day spent in space), and the lowest rates in the 1990s and the period from 2010 until the present (no deaths). The most dangerous phases of spaceflight are launch, landing and staying in orbit. Altogether, n = 12 incidents (incident rate per spaceflight: 0.04) and one accident (accident rate: 0.003) during launch have been reported, n = 9 incidents (incident rate: 0.03) and two accidents (accident rate: 0.006) have been reported during landing and n = 10 incidents (incident rate: 0.03) have been reported in orbit. Discussion: Manned spaceflight over the last six decades has become significantly safer. Since 2003, no astronaut fatality has been reported. With greater international cooperation and maintaining of the International Space Station (ISS), the number of manned spaceflights and days spent in space has constantly increased, with constantly lower rates of incidents and accidents.