Search

Background: Allocation of scarce organs for transplantation is ethically challenging. Artificial intelligence (AI) has been proposed to assist in liver allocation, however the ethics of this remains unexplored and the view of the public unknown. The aim of this paper was to assess public attitudes on whether AI should be used in liver allocation and how it should be implemented. Methods: We first introduce some potential ethical issues concerning AI in liver allocation, before analysing a pilot survey including online responses from 172 UK laypeople, recruited through Prolific Academic. Findings: Most participants found AI in liver allocation acceptable (69.2%) and would not be less likely to donate their organs if AI was used in allocation (72.7%). Respondents thought AI was more likely to be consistent and less biased compared to humans, although were concerned about the "dehumanisation of healthcare" and whether AI could consider important nuances in allocation decisions. Participants valued accuracy, impartiality, and consistency in a decision-maker, more than interpretability and empathy. Respondents were split on whether AI should be trained on previous decisions or programmed with specific objectives. Whether allocation decisions were made by transplant committee or AI, participants valued consideration of urgency, survival likelihood, life years gained, age, future medication compliance, quality of life, future alcohol use and past alcohol use. On the other hand, the majority thought the following factors were not relevant to prioritisation: past crime, future crime, future societal contribution, social disadvantage, and gender. Conclusions: There are good reasons to use AI in liver allocation, and our sample of participants appeared to support its use. If confirmed, this support would give democratic legitimacy to the use of AI in this context and reduce the risk that donation rates could be affected negatively. Our findings on specific ethical concerns also identify potential expectations and reservations laypeople have regarding AI in this area, which can inform how AI in liver allocation could be best implemented. [ABSTRACT FROM AUTHOR]

Much of the knowledge and information needed for enabling high-quality clinical research is stored in free-text format. Natural language processing (NLP) has been used to extract information from these sources at scale for several decades. This paper aims to present a comprehensive review of clinical NLP for the past 15 years in the UK to identify the community, depict its evolution, analyse methodologies and applications, and identify the main barriers. We collect a dataset of clinical NLP projects (n = 94; £ = 41.97 m) funded by UK funders or the European Union's funding programmes. Additionally, we extract details on 9 funders, 137 organisations, 139 persons and 431 research papers. Networks are created from timestamped data interlinking all entities, and network analysis is subsequently applied to generate insights. 431 publications are identified as part of a literature review, of which 107 are eligible for final analysis. Results show, not surprisingly, clinical NLP in the UK has increased substantially in the last 15 years: the total budget in the period of 2019–2022 was 80 times that of 2007–2010. However, the effort is required to deepen areas such as disease (sub-)phenotyping and broaden application domains. There is also a need to improve links between academia and industry and enable deployments in real-world settings for the realisation of clinical NLP's great potential in care delivery. The major barriers include research and development access to hospital data, lack of capable computational resources in the right places, the scarcity of labelled data and barriers to sharing of pretrained models. [ABSTRACT FROM AUTHOR]

This paper determines the assessment of publications submitted to the UK research evaluation carried out in 2014, the REF, which would have resulted if they had been assessed with the bibliometric algorithm used by the Italian evaluation agency, ANVUR, for its evaluation of the research of Italian universities. We find extremely high correlations between the two assessment approaches. [ABSTRACT FROM AUTHOR]

With the development of genome-wide association studies, how to gain information from a large scale of data has become an issue of common concern, since traditional methods are not fully developed to solve problems such as identifying loci-to-loci interactions (also known as epistasis). Previous epistatic studies mainly focused on local information with a single outcome (phenotype), while in this paper, we developed a two-stage global search algorithm, Greedy Equivalence Search with Local Modification (GESLM), to implement a global search of directed acyclic graph in order to identify genome-wide epistatic interactions with multiple outcome variables (phenotypes) in a case–control design. GESLM integrates the advantages of score-based methods and constraint-based methods to learn the phenotype-related Bayesian network and is powerful and robust to find the interaction structures that display both genetic associations with phenotypes and gene interactions. We compared GESLM with some common phenotype-related loci detecting methods in simulation studies. The results showed that our method improved the accuracy and efficiency compared with others, especially in an unbalanced case–control study. Besides, its application on the UK Biobank dataset suggested that our algorithm has great performance when handling genome-wide association data with more than one phenotype. [ABSTRACT FROM AUTHOR]

Background Evidence suggests a correlation of blood pressure (BP) level with presence of diabetic microvascular complications (DMCs), but the effect of BP on DMCs incidence is not well-established. We aimed to explore the associations between BP and DMCs (diabetic retinopathy, diabetic kidney disease, and diabetic neuropathy) risk in participants with diabetes. Methods This study included 23030 participants, free of any DMCs at baseline, from the UK Biobank. We applied multivariable-adjusted Cox regression models to estimate BP-DMCs association and constructed BP genetic risk scores (GRSs) to test their association with DMCs phenotypes. Differences in incidences of DMCs were also compared between the 2017 ACC/AHA and JNC 7 guidelines (traditional criteria) of hypertension. Results Compared to systolic blood pressure (SBP)<120 mm Hg, participants with SBP≥160 mm Hg had a hazard ratio (HR) of 1.50 (95% confidence interval (CI)=1.09, 2.06) for DMCs. Similarly, DMCs risk increased by 9% for every 10 mm Hg of higher SBP at baseline (95% CI=1.04, 1.13). The highest tercile SBP GRS was associated with 32% higher DMCs risk (95% CI=1.11, 1.56) compared to the lowest tercile. We found no significant differences in DMCs incidence between JNC 7 and 2017 ACC/AHA guidelines. Conclusions Genetic and epidemiological evidence suggests participants with higher SBP had an increased risk of DMCs, but hypertension defined by 2017 ACC/AHA guidelines may not impact DMCs incidence compared with JNC 7 criteria, contributing to the care and prevention of DMCs. [ABSTRACT FROM AUTHOR]

Wearable inertial measurement units (IMUs) are being used to quantify gait characteristics that are associated with increased fall risk, but the current limitation is the lack of contextual information that would clarify IMU data. Use of wearable video-based cameras would provide a comprehensive understanding of an individual's habitual fall risk, adding context to clarify abnormal IMU data. Generally, there is taboo when suggesting the use of wearable cameras to capture real-world video, clinical and patient apprehension due to ethical and privacy concerns. This perspective proposes that routine use of wearable cameras could be realized within digital medicine through AI-based computer vision models to obfuscate/blur/shade sensitive information while preserving helpful contextual information for a comprehensive patient assessment. Specifically, no person sees the raw video data to understand context, rather AI interprets the raw video data first to blur sensitive objects and uphold privacy. That may be more routinely achieved than one imagines as contemporary resources exist. Here, to showcase/display the potential an exemplar model is suggested via off-the-shelf methods to detect and blur sensitive objects (e.g., people) with an accuracy of 88%. Here, the benefit of the proposed approach includes a more comprehensive understanding of an individual's free-living fall risk (from free-living IMU-based gait) without compromising privacy. More generally, the video and AI approach could be used beyond fall risk to better inform habitual experiences and challenges across a range of clinical cohorts. Medicine is becoming more receptive to wearables as a helpful toolbox, camera-based devices should be plausible instruments. [ABSTRACT FROM AUTHOR]

This paper aims to demonstrate how knowledge acquired from occupants' manual window operations can be implemented into BMS automated window operation algorithms. Ten single-occupant offices were selected in a university building in the UK. More than 28,000 hourly data points on indoor and outdoor temperature and open window area (OWA) were analysed from 2015 to 2020. The BMS had adopted nine different automated window operation algorithms during the 5 years. The automated window algorithms could be manually overridden by the office occupants. Automated algorithms were compared against manual window operations. The results showed that the slope and gradient of the regression lines for occupants' manual window operations are smaller than automated operations. OWA of automated window operations increased 20% per 1 °C increase in indoor temperature, however, occupants opened windows 6–8% per 1 °C increase. Occupants react slower to temperature changes than assumed by BMS, which could be considered in BMS automated window operations. • BMS automated window algorithms operate based on indoor temperature in this study. • Automated window operations by BMS impact occupants' manual window operations. • Occupants open windows at higher temperatures on the third floor than first floor. • BMSs need to consider building-related differences and seasonal changes. • A BMS system architecture is proposed to reduce the behaviour gap. [ABSTRACT FROM AUTHOR]

In August 2020, the UK government and regulation body Ofqual replaced school examinations with automatically computed A Level grades in England and Wales. This algorithm factored in school attainment in each subject over the previous three years. Government officials initially stated that the algorithm was used to combat grade inflation. After public outcry, teacher assessment grades used instead. Views concerning who was to blame for this scandal were expressed on the social media website Twitter. While previous work used NLP-based opinion mining computational linguistic tools to analyse this discourse, shortcomings included accuracy issues, difficulties in interpretation and limited conclusions on who authors blamed. Thus, we chose to complement this research by analysing 18,239 tweets relating to the A Level algorithm using Corpus Linguistics (CL) and Critical Discourse Analysis (CDA), underpinned by social actor representation. We examined how blame was attributed to different entities who were presented as social actors or having social agency. Through analysing transitivity in this discourse, we found the algorithm itself, the UK government and Ofqual were all implicated as potentially responsible as social actors through active agency, agency metaphor possession and instances of passive constructions. According to our results, students were found to have limited blame through the same analysis. We discuss how this builds upon existing research where the algorithm is implicated and how such a wide range of constructions obscure blame. Methodologically, we demonstrated that CL and CDA complement existing NLP-based computational linguistic tools in researching the 2020 A Level algorithm; however, there is further scope for how these approaches can be used in an iterative manner. [ABSTRACT FROM AUTHOR]

During the outbreak of a disease caused by a pathogen with unknown characteristics, the uncertainty of its progression parameters can be reduced by devising methods that, based on rational assumptions, exploit available information to provide actionable insights. In this study, performed a few (~6) weeks into the outbreak of COVID-19 (caused by SARS-CoV-2), one of the most important disease parameters, the average time-to-recovery, was calculated using data publicly available on the internet (daily reported cases of confirmed infections, deaths, and recoveries), and fed into an algorithm that matches confirmed cases with deaths and recoveries. Unmatched cases were adjusted based on the matched cases calculation. The mean time-to-recovery, calculated from all globally reported cases, was found to be 18.01 days (SD 3.31 days) for the matched cases and 18.29 days (SD 2.73 days) taking into consideration the adjusted unmatched cases as well. The proposed method used limited data and provided experimental results in the same region as clinical studies published several months later. This indicates that the proposed method, combined with expert knowledge and informed calculated assumptions, could provide a meaningful calculated average time-to-recovery figure, which can be used as an evidence-based estimation to support containment and mitigation policy decisions, even at the very early stages of an outbreak. [ABSTRACT FROM AUTHOR]

Cardiorespiratory fitness is an established predictor of metabolic disease and mortality. Fitness is directly measured as maximal oxygen consumption (VO2max), or indirectly assessed using heart rate responses to standard exercise tests. However, such testing is costly and burdensome because it requires specialized equipment such as treadmills and oxygen masks, limiting its utility. Modern wearables capture dynamic real-world data which could improve fitness prediction. In this work, we design algorithms and models that convert raw wearable sensor data into cardiorespiratory fitness estimates. We validate these estimates' ability to capture fitness profiles in free-living conditions using the Fenland Study (N=11,059), along with its longitudinal cohort (N = 2675), and a third external cohort using the UK Biobank Validation Study (N = 181) who underwent maximal VO2max testing, the gold standard measurement of fitness. Our results show that the combination of wearables and other biomarkers as inputs to neural networks yields a strong correlation to ground truth in a holdout sample (r = 0.82, 95CI 0.80–0.83), outperforming other approaches and models and detects fitness change over time (e.g., after 7 years). We also show how the model's latent space can be used for fitness-aware patient subtyping paving the way to scalable interventions and personalized trial recruitment. These results demonstrate the value of wearables for fitness estimation that today can be measured only with laboratory tests. [ABSTRACT FROM AUTHOR]

Objectives X-Linked hypophosphataemic rickets (XLH) is a rare multi-systemic disease of mineral homeostasis that has a prominent skeletal phenotype. The aim of this study was to describe additional comorbidities in XLH patients compared with general population controls. Methods The Clinical Practice Research Datalink (CPRD) GOLD was used to identify a cohort of XLH patients (1995–2016), along with a non-XLH cohort matched (1 : 4) on age, sex and GP practice. Using the CALIBER portal, phenotyping algorithms were used to identify the first diagnosis (and associated age) of 273 comorbid conditions during patient follow-up. Fifteen major disease categories were used and the proportion of patients having ≥1 diagnosis was compared between cohorts for each category and condition. Main analyses were repeated according to the Index of Multiple Deprivation (IMD). Results There were 64 and 256 patients in the XLH and non-XLH cohorts, respectively. There was increased prevalence of endocrine [OR 3.46 (95% CI: 1.44, 8.31)] and neurological [OR 3.01 (95% CI: 1.41, 6.44)] disorders among XLH patients. Across all specific comorbidities, four were at least twice as likely to be present in XLH cases, but only depression met the Bonferroni threshold: OR 2.95 (95% CI: 1.47, 5.92). Distribution of IMD among XLH cases indicated greater deprivation than the general population. Conclusion We describe a higher risk of mental illness in XLH patients compared with matched controls, and greater than expected deprivation. These findings may have implications for clinical practice guidelines and decisions around health and social care provision for these patients. [ABSTRACT FROM AUTHOR]

River-level estimation is a critical task required for the understanding of flood events and is often complicated by the scarcity of available data. Recent studies have proposed to take advantage of large networks of river-camera images to estimate river levels but, currently, the utility of this approach remains limited as it requires a large amount of manual intervention (ground topographic surveys and water image annotation). We have developed an approach using an automated water semantic segmentation method to ease the process of river-level estimation from river-camera images. Our method is based on the application of a transfer learning methodology to deep semantic neural networks designed for water segmentation. Using datasets of image series extracted from four river cameras and manually annotated for the observation of a flood event on the rivers Severn and Avon, UK (21 November–5 December 2012), we show that this algorithm is able to automate the annotation process with an accuracy greater than 91%. Then, we apply our approach to year-long image series from the same cameras observing the rivers Severn and Avon (from 1 June 2019 to 31 May 2020) and compare the results with nearby river-gauge measurements. Given the high correlation (Pearson's correlation coefficient >0.94) between these results and the river-gauge measurements, it is clear that our approach to automation of the water segmentation on river-camera images could allow for straightforward, inexpensive observation of flood events, especially at ungauged locations. [ABSTRACT FROM AUTHOR]

Background: Cardiovascular death is a common outcome in population-based studies about new healthcare interventions or treatments, such as new prescription medications. Vital statistics registration systems are often the preferred source of information about cause-specific mortality because they capture verified information about the deceased, but they may not always be accessible for linkage with other sources of population-based data. We assessed the validity of an algorithm applied to administrative health records for identifying cardiovascular deaths in population-based data.Methods: Administrative health records were from an existing multi-database cohort study about sodium-glucose cotransporter-2 (SGLT2) inhibitors, a new class of antidiabetic medications. Data were from 2013 to 2018 for five Canadian provinces (Alberta, British Columbia, Manitoba, Ontario, Quebec) and the United Kingdom (UK) Clinical Practice Research Datalink (CPRD). The cardiovascular mortality algorithm was based on in-hospital cardiovascular deaths identified from diagnosis codes and select out-of-hospital deaths. Sensitivity, specificity, and positive and negative predictive values (PPV, NPV) were calculated for the cardiovascular mortality algorithm using vital statistics registrations as the reference standard. Overall and stratified estimates and 95% confidence intervals (CIs) were computed; the latter were produced by site, location of death, sex, and age.Results: The cohort included 20,607 individuals (58.3% male; 77.2% ≥70 years). When compared to vital statistics registrations, the cardiovascular mortality algorithm had overall sensitivity of 64.8% (95% CI 63.6, 66.0); site-specific estimates ranged from 54.8 to 87.3%. Overall specificity was 74.9% (95% CI 74.1, 75.6) and overall PPV was 54.5% (95% CI 53.7, 55.3), while site-specific PPV ranged from 33.9 to 72.8%. The cardiovascular mortality algorithm had sensitivity of 57.1% (95% CI 55.4, 58.8) for in-hospital deaths and 72.3% (95% CI 70.8, 73.9) for out-of-hospital deaths; specificity was 88.8% (95% CI 88.1, 89.5) for in-hospital deaths and 58.5% (95% CI 57.3, 59.7) for out-of-hospital deaths.Conclusions: A cardiovascular mortality algorithm applied to administrative health records had moderate validity when compared to vital statistics data. Substantial variation existed across study sites representing different geographic locations and two healthcare systems. These variations may reflect different diagnostic coding practices and healthcare utilization patterns. [ABSTRACT FROM AUTHOR]