Search

The study explores the application of automated machine learning (AutoML) using the MIMIC-IV-ED database to enhance decision-making in emergency department (ED) triage. We developed a predictive model that utilizes triage data to forecast hospital admissions, aiming to support medical staff by providing an advanced decision-support system. The model, powered by H2O.ai’s AutoML platform, was trained on approximately 280,000 preprocessed records from the Beth Israel Deaconess Medical Center collected between 2011 and 2019. The selected Gradient Boosting Machine (GBM) model demonstrated an AUC ROC of 0.8256, indicating its efficacy in predicting patient dispositions. Key variables such as acuity and waiting hours were identified as significant predictors, emphasizing the model’s capability to integrate critical triage metrics into its predictions. However, challenges related to the complexity and heterogeneity of medical data, privacy concerns, and the need for model interpretability were addressed through the incorporation of Explainable AI (XAI) techniques. These techniques ensure the transparency of the predictive processes, fostering trust and facilitating ethical AI use in clinical settings. Future work will focus on external validation and expanding the model to include a broader array of variables from diverse healthcare environments, enhancing the model’s utility and applicability in global emergency care contexts.

The purpose of this study is to evaluate the institutional and tax environment for businesses in 21 countries of the European Union (EU), using a multi-criteria analysis and, in particular, the PROMETHEE II method, based on thirteen evaluation criteria and the ranking of the European economies. Through the research, the importance of factors related to the efficiency of the tax administration, the tax capacity of the country, the growth rate of the economy, the quality of institutions, the level of corruption, the tax burden, the time of tax compliance and the political stability of the country is studied, in the formation of an attractive environment for the development of entrepreneurship in the EU economies. Moreover, the results of the analysis revealed the existence of significant fiscal and institutional differences between the EU economies under investigation, regarding the development of entrepreneurship. These differences are due to the size of the economy, the institutions, the history, the economic crises of each country and a wide range of dissimilar policy recommendations, which reflect asymmetric policy responses. The study concludes by offering theoretical and political recommendations to improve the institutional and fiscal environment in EU economies so that entrepreneurship can play an important role in improving the economic and social situation of a country.

(1) Background: Predictive modeling is becoming increasingly relevant in healthcare, aiding in clinical decision making and improving patient outcomes. However, many of the most potent predictive models, such as deep learning algorithms, are inherently opaque, and their decisions are challenging to interpret. This study addresses this challenge by employing Shapley Additive Explanations (SHAP) to facilitate model interpretability while maintaining prediction accuracy. (2) Methods: We utilized Gradient Boosting Machines (GBMs) to predict patient outcomes in an emergency department setting, with a focus on model transparency to ensure actionable insights. (3) Results: Our analysis identifies “Acuity”, “Hours”, and “Age” as critical predictive features. We provide a detailed exploration of their intricate interactions and effects on the model’s predictions. The SHAP summary plots highlight that “Acuity” has the highest impact on predictions, followed by “Hours” and “Age”. Dependence plots further reveal that higher acuity levels and longer hours are associated with poorer patient outcomes, while age shows a non-linear relationship with outcomes. Additionally, SHAP interaction values uncover that the interaction between “Acuity” and “Hours” significantly influences predictions. (4) Conclusions: We employed force plots for individual-level interpretation, aligning with the current shift toward personalized medicine. This research highlights the potential of combining machine learning’s predictive power with interpretability, providing a promising route concerning a data-driven, evidence-based healthcare future.

This study proposes an integrated methodology for the assessment and mapping of quality of life (QoL) among European Union member states in the period before and after the pandemic crisis of COVID-19. The assessment of quality of life was based on the development of composite criteria and Geographical Information Systems or GIS technology, using variables that assess quality of life. The composite criteria relate to the socioeconomic environment, employment conditions, economic conditions and health services. Each criterion was evaluated by a set of variables, and each variable was weighted based on relevant research by Greek experts. Criteria were also weighted and combined to assess overall quality of life. The methodology was applied in 27 EU member countries, and mapping led to the identification of countries with low and high quality of life. The results showed a change in the level of overall quality of life in the EU countries before and after the pandemic period, although on a limited scale, since there is a slight reclassification of the countries’ positions. The analysis also revealed the highest level of quality of life in four EU countries [Sweden, Denmark, the Netherlands and Luxembourg] that show an increased GDP per capita, combining a low level of arrears and a low level of inability to make ends meet, whereas four countries showed the lowest level of quality of life [Greece, Bulgaria, Romania and Croatia] in both periods.

This research ventures into the critical public health challenge of childhood obesity by exploring the dynamic interplay between psychological well-being and Body Mass Index (BMI) throughout various developmental stages of childhood. It delves into how emotional regulation, attachment dynamics, and social relationships correlate with obesity from early childhood to adolescence. Highlighting key findings, such as the negative correlation between psychological resilience and higher BMI in young children, the impact of social relationships on obesity risk during pre-adolescence, and the link between adaptive emotional strategies and higher BMI in adolescents, this study brings to the fore the nuanced relationship between psychological factors and obesity. Psychological metrics in this study were obtained via referenced questionnaires, leading up to the utilization of the interdisciplinary process of bioinformatics. Utilizing the interdisciplinary process of bioinformatics, this research synergizes psychometric and biomedical data to unearth psychological markers critical for crafting targeted, age-appropriate interventions. This study advocates for a holistic healthcare approach, emphasizing the integration of psychological support within obesity prevention and management strategies, thereby underscoring the indispensable role of psychological factors in the fight against childhood obesity. The application of bioinformatics methods to analyze complex datasets demonstrates how collaboration across medical specialties can enrich our understanding and response to childhood obesity, contributing significantly to the development of comprehensive, bioinformatics-enhanced healthcare solutions.

Innovation systems consist of different organisations from the quadruple helix, as well as the interactions and linkages between them. Smart technologies and ICT play a key role in the efficiency of systems. At the same time, regional scale is considered crucial for studying innovation in systems. However, the lack of many important data at the regional level compounds the efforts to study them. The paper proposes a novel methodological approach to the regionalisation of national-level indicators in order to address this issue. This is based on the model fit approach, using regressions to “regionalise” national-level indicators based on similar indicators that are available. The approach is tested on the data for Greek NUTS 2 regions and produces regional-level estimates for four innovation indicators, based on four available indicators that are found to be strongly correlated to them. However, the same approach can be used for any EU country or the whole of the EU. The results, their prospects for future research, and potential applications are considered. Overall, the availability of regional-level indicators is considered crucial for the formulation of impactful development policies.

IntroductionElectronic Health Records (EHRs) are essential data structures, enabling the sharing of valuable medical care information for a diverse patient population and being reused as input to predictive models for clinical research. However, issues such as the heterogeneity of EHR data and the potential compromisation of patient privacy inhibit the secondary use of EHR data in clinical research.ObjectivesThis study aims to present the main elements of the MODELHealth project implementation and the evaluation method that was followed to assess the efficiency of its mechanism.MethodsThe MODELHealth project was implemented as an Extract-Transform-Load system that collects data from the hospital databases, performs harmonization to the HL7 FHIR standard and anonymization using the k-anonymity method, before loading the transformed data to a central repository. The integrity of the anonymization process was validated by developing a database query tool. The information loss occurring due to the anonymization was estimated with the metrics of generalized information loss, discernibility and average equivalence class size for various values of k.ResultsThe average values of generalized information loss, discernibility and average equivalence class size obtained across all tested datasets and k values were 0.008473 ± 0.006216252886, 115,145,464.3 ± 79,724,196.11 and 12.1346 ± 6.76096647, correspondingly. The values of those metrics appear correlated with factors such as the k value and the dataset characteristics, as expected.ConclusionThe experimental results of the study demonstrate that it is feasible to perform effective harmonization and anonymization on EHR data while preserving essential patient information.

Several empirical papers have studied the negative effects of taxation on economic activities and especially on economic growth. There are several ways to estimate the negative effects of several kinds of taxation. The majority of authors analyses how tax on income influences economic growth and how taxes on corporate income affect economic growth. Other articles examine how to design an effective tax policy for economic recovery and growth. The analysis focuses on the impact of tax structures to long- run growth. This article examines how distortionary taxation and non-distortionary taxation influence growth. Specifically, in the empirical analysis, this paper uses different kinds of taxes such as taxes on income. All fiscal variables are expressed as percentage of GDP and were extracted from IMF database. Empirical analysis used time series for Greece from 1973 until 2018.

Mental workload has a major effect on the individual’s performance in most real-world tasks, which can lead to significant errors in critical operations. On this premise, the analysis and assessment of mental workload attain high research interest in both the fields of Neuroergonomics and Neuroscience. In this work, we implemented an EEG experimental design consisting of two distinct mental tasks (mental arithmetic task, n-back task), each with two conditions of complexity (low and high) to investigate the task-related and task-unrelated workload effects. Since mental workload is an intricate phenomenon involving multiple brain areas, we performed a graph theoretical analysis estimating the Phase Locking Index (PLI) in four frequency bands (delta, theta, alpha, beta). The brainwave-dependent network results show statistically significant reductions in clustering coefficient, characteristic path length, and small-worldness metrics with higher workload in both tasks across several bands. Moreover, functional connectivity analysis indicates a task-independent fashion of the brain topological re-organization with increasing mental load. These results revealed how the brain network is re-organized with increasing mental workload in a task-independent way. Finally, the network metrics were used as classification features, leading to high performance in workload level discrimination.

The detection of mental fatigue is an important issue in the nascent field of neuroergonomics. Although machine learning approaches and especially deep learning designs have constantly demonstrated their efficiency to automatically detect critical features from raw data, the computational resources for training and predictions are usually very demanding. In this work, we propose a shallow convolutional neural network, with three convolutional layers, for fatigue detection using electroencephalogram (EEG) data that can alleviate the computational burden and provide fast mental fatigue detection. As such, a deep learning model was created utilizing time-frequency domain features, extracted with Morlet wavelet analysis. These features, combined with the higher-level characteristics learnt by the model, resulted in a resilient solution, able to attain very high prediction accuracy (97%), while reducing training time and computing costs. Moreover, by incorporating a subsequent SHAP values analysis on the characteristics that contributed in the model creation, indications of low frequency (theta and alpha band) brain wave characteristics were indicated as prominent mental fatigue detectors.

The object of this paper was the application of machine learning to a clinical dataset that was anonymized using the Mondrian algorithm. (1) Background: The preservation of patient privacy is a necessity rising from the increasing digitization of health data; however, the effect of data anonymization on the performance of machine learning models remains to be explored. (2) Methods: The original EHR derived dataset was subjected to anonymization by applying the Mondrian algorithm for various k values and quasi identifier (QI) set attributes. The logistic regression, decision trees, k-nearest neighbors, Gaussian naive Bayes and support vector machine models were applied to the different dataset versions. (3) Results: The classifiers demonstrated different degrees of resilience to the anonymization, with the decision tree and the KNN models showing remarkably stable performance, as opposed to the Gaussian naïve Bayes model. The choice of the QI set attributes and the generalized information loss value played a more important role than the size of the QI set or the k value. (4) Conclusions: Data anonymization can reduce the performance of certain machine learning models, although the appropriate selection of classifier and parameter values can mitigate this effect.

As life expectancy increases, it is imperative that the elderly take advantage of the benefits of technology to remain active and independent. Mobile health applications are widely used nowadays as they promote a healthy lifestyle and self-management of diseases, opening new horizons in the interactive health service delivery. However, adapting these applications to the needs and requirements of the elderly is still a challenge. This article presents a smartphone application that is part of a multifactorial intervention to support older people with balance disorders. The application aims to enable users to self-evaluate their activity and progress, to communicate with each other and, through strategically selected motivational features, to engage with the system with undiminished interest for a long period of time. Mock-up interfaces were evaluated in semi-structured focus groups and interviews that were performed across three European countries. Further evaluation in the form of four pilot studies with 160 participants will be performed and qualitative and quantitative measures will be used to process the feedback about the use of the application.

In the history of human kind in all ethnic groups and communities there have been individuals with physical or mental disabilities. Until the 18th century not only these individuals did not receive any support or care by their community but they were rather treated as inferior with a diminishing way. More or less the same attitude occurs to the ancient Greek world, where old age, limited psychomotor skills and potential disability were perceived as evidence of deprivation of the grace of gods. A minor acknowledgment of individuals with disabilities started from Athens, since there were several benefits introduced on behalf of the disabled for the very first time in the history of human kind. In the ancient times Greeks made use of physical exercise as a mean for therapeutic as well as prevention purposes against various diseases. Thus, a great number of distinguished physicians such as Herodikos and his student Hippocrates recommended physical exercise as the ideal treatment of several health related problems.

Cross-linked polyethylene (XLPE) and silicone rubber (SiR) samples were subjected to a high-voltage AC stress plane–plane configuration and inclined plane test, respectively. The voltage was applied such that discharge was observed across the surface of the XLPE test sample for several hours and for visible damage to occur on SiR samples also after several hours. Selected stressed samples together with virgin samples from the same manufactured batch were tested using nuclear magnetic resonance (NMR) spectroscopy. Specifically, ^1H NMR spin–lattice (T(1)) and spin–spin (T(2)) relaxation time measurements were employed to examine potential changes in the chemical bonding of undamaged and damaged XLPE and SiR samples. Preliminary results show that there may be a moderate increase in the T(1) and T(2) values of the damaged samples in comparison with the undamaged ones. This raises the possibility that NMR can be a useful additional experimental tool in characterising material degradation.

Introduction Discovering suitable datasets is an important part of health research, particularly for projects using linked and/or cohort data, but with the proliferation of so many national and international initiatives, it is becoming increasingly difficult for research teams to locate real world datasets that are most relevant to their project objectives. Objectives and Approach To assist researchers in this, a bibliographic data analysis platform, bibInsight, is being developed which aims to identify potentially useful datasets, among other information, from large numbers of publications. A search dataset on the broad topic of "dementia" was obtained from PubMed and abstracts mentioning cohort names were identified. These helped in determining the context within which cohort data are mentioned. Terms were also informed by specialist knowledge from the European Medical Information Framework (EMIF) Alzheimer's Disease project, and were used to train a one-class classifier that identified their structure. Results The initial PubMed search resulted in 129,040 articles, with 1961 identified cohorts. From these, the classifier was able to identify 1,588 as containing cohort information, and 373 as not, a precision of approximately 80\%. The terms identified by the classifier included "annual", "longitudinal", and "prospective", suggesting re-contact. Further analysis of these 1,588 abstracts can find those most relevant to specific topics, co-authorship collaborations (suggesting data sharing agreements are already in place), and mention of linkage to other datasets. These can be done through keyword extraction, longitudinal data mining methods, linking author affiliations to institutions through the Global Resource Information Database (GRID), and graph databases, all of which have been built into the bibInsight platform. Conclusion/Implications `This approach shows promise in facilitating research by making literature searches more efficient. A key issue is working with unstructured text data which requires additional standardisation for analysis. Data reuse is a growing area, and publications can assist this, but better reporting would be beneficial.

Introduction Routine data sources are increasingly being explored as an alternative to surveys for governmental censuses. Censuses are useful, but only capture one snapshot in time. For research, changes in state over time can be important. This is particularly true for households, where changes in both composition and location are routine. Objectives and Approach The Secure Anonymised Information Linkage (SAIL) databank at Swansea University holds anonymised address data tied to Welsh General Practice (GP) registrations. The dataset has the individual’s unique identifier (ALF), address identifier (RALF), and moving in and moving out dates. This can be securely linked to other demographic and health data. Using this data, a finite, deterministic household model was developed. This model implements a set of rules for establishing, advancing, and terminating households. The rules cover different types of households, including transient and multi-generational, but not communal residences such as student halls or residential care homes. Results The model was implemented in Python, and operates over finite moving in/out transactions undertaken by actors (ALFs) on specific properties (RALFs). The roles also help to diminish data issues, such as infants moving into an address before their parents. The model creates and updates histories of households for each RALF. This allows researchers to identify unstable versus stable households, or to look at possible environmental health factors over time. The script also provides a running count of the total number of residents at each RALF at any given time. This makes it useful for time-specific research, or for linking to censuses. Conclusion/Implications The ability to have a longitudinal record of households can provide additional detail to research which can be costly to obtain otherwise. This method shows promise as a useful tool for linked population data research. Further validation against surveys, and work to identify and model communal residences is needed.

Background: Up to half of patients with dementia may not receive a formal diagnosis, limiting access to appropriate services. It is hypothesised that it may be possible to identify undiagnosed dementia from a profile of symptoms recorded in routine clinical practice. Aim: The aim of this study is to develop a machine learning-based model that could be used in general practice to detect dementia from routinely collected NHS data. The model would be a useful tool for identifying people who may be living with dementia but have not been formally diagnosed. Design & setting: The study involved a case-control design and analysis of primary care data routinely collected over a 2-year period. Dementia diagnosed during the study period was compared to no diagnosis of dementia during the same period using pseudonymised routinely collected primary care clinical data. Method: Routinely collected Read-encoded data were obtained from 18 consenting GP surgeries across Devon, for 26 483 patients aged >65 years. The authors determined Read codes assigned to patients that may contribute to dementia risk. These codes were used as features to train a machine-learning classification model to identify patients that may have underlying dementia. Results: The model obtained sensitivity and specificity values of 84.47% and 86.67%, respectively. Conclusion: The results show that routinely collected primary care data may be used to identify undiagnosed dementia. The methodology is promising and, if successfully developed and deployed, may help to increase dementia diagnosis in primary care.

Background Discovering suitable datasets is an important part of health research, particularly for projects working with cohort data, but with the proliferation of so many national and international initiatives, it is becoming increasingly difficult for research teams to locate real world datasets that are most relevant to their project objectives. Methods To assist researchers in this, we developed bibInsight, a data analysis platform to identify potentially useful data sources and more generally enable large scale research over bibliographical datasets. Data source names were identified from a broad, topic specific literature search. Context-specific terms like “annual”, “longitudinal”, and “prospective” were used to train a classifier that identified potential datasets. Results The classifier was able to identify 1588 of 1961 abstracts as containing cohort-relevant information: a precision of approximately 80%. Further analysis such as topic analysis, geographical mapping, and collaboration networks can refine and prioritise the search results to determine the most relevant data source(s) for a research project. Conclusions A very large amount of information, including data source description and use, remains unexploited in unstructured bibliographical datasets. Here, we used a thematic search to provide a more manageable starting point towards locating disease specific datasets.

ABSTRACT Objectives The poster will showcase the Training & Capacity Building Programme (TCBP) established by the Farr network that is available to PhD students and Early Career Researchers (ECRs) both from within and outside the network. Approach The authors are using a mixed-methods approach to present the current state of the art that has motivated the structure of the TCBP and how the programme has been received by students, academic and industry leaders. Since 2014, the Farr network education leads have worked on a unique education programme that aims to equip students with key professional and methodological skills required by the roles they are likely to take in their future careers. The content of the programme has been informed by current research in the fields of data linkage and analysis of big datasets as well as the immediate experience of researchers, and healthcare and industry leads working actively in the field. We have focussed on developing a community of practice through the provision of an environment that has enabled students to share experiences and knowledge and to start building their own networks of collaboration across the Farr centres. Results The Farr network of PhD students comprises of approximately 66 students from a variety of backgrounds including bioinformatics, computer science, epidemiology, mathematics, psychology and statistics. The Farr network hosted 2 main events in 2015 that provided the foundation for 34 PhD students to come together to develop skills in written and oral communication, Public and Patient Involvement (PPI), problem-solving, requirements gathering, data visualisation and spatial data analysis. Formal feedback collected from each event suggests that the programme is received well by students who see it as a set of activities that complement their specialised PhD related education. Feedback from supervisors and future employers also suggests that this programme has facilitated the development of skills that are useful for their PhD and employment. Conclusion The Farr network has laid the foundation for the community of practice around the analysis of real-world health datasets across the UK, and will continue to cultivate this through the enrichment of the programme.

ABSTRACT Objectives The main objective of this poster is to present a pilot project in determining emerging population health themes and identifying key research-enabling datasets ahead of time. At present, large-scale databanks, such as the Secure Anonymised Information Linkage (SAIL) Databank at Swansea University Medical School, already manage large quantities of health and administrative linked datasets. While these datasets are valuable for research purposes, complementary datasets may be required by collaborating researchers to answer detailed population health research questions. Dataset acquisition can take several years, which is a serious delay to a project with time-limited funding. The ability to pre-emptively acquire datasets so that these are ready for use before a researcher requests them would obviously be beneficial. However, a recent study conducted by the Farr Cipher team at Swansea University identified over 800 health and administrative datasets in Wales alone. With limited resources such as available funding and time, which of these datasets is worth its effort in acquiring? Approach Bibliometrics has long been a means of measuring the impact of papers on the wider academic community. Lately, the focus of analyses has been extended to include the topics, authorship and citations of the publications. Existing bibliometric data mining techniques suggest that it is possible to identify emerging topic trends and through this assist in prioritising dataset identification and acquisition. The project explored mining available literature through bibliometric analysis in order to predict emerging trends and through these identify potentially relevant and valuable datasets for acquisition on behalf of the Dementias Platform UK (DPUK). Literature searches were conducted for papers published on the topic of “dementia” over the last 20 years. Additional keywords and topics were extracted to identify emerging areas of research and clinical interest. These were then compared against an existing list of over 800 Welsh datasets currently not held in SAIL. Results Results focus on: • Using bibliometric methods in the context of DPUK cohort publications • Identifying emerging trends in the field of dementia research. • Identifying and prioritising datasets which might be useful for the SAIL Databank to acquire