A Preliminary Study on the Abnormal Deaths and Work Burden of Chinese Physicians: A Mixed Method Analysis and Implications for Smart Hospital Management

Background: The population of Chinese physicians is frequently threatened by abnormal death, including death by overwork or homicide. This is not only a health problem, but also a social problem that has attracted the attention of both hospitals and the government.Objective: This study aims to analyze the characteristics of abnormal death in physicians in Chinese hospitals from 2007 to 2020 and to investigate the relationship between abnormal death and physician workload, in order to provide information for policy makers and request improvement technologies.Methods: A mixed research method was used. In order to ensure accuracy and completeness, a relatively comprehensive search was conducted using multiple heterogeneous data sources on the abnormal death of physicians in Chinese hospitals from 2007 to 2020. The collected cases were then descriptively analyzed using the work-related overwork death risk concept framework and the deductive grounded theory approach. In addition, the workload of physicians was calculated between 2007 and 2019 based on three important workload indicators.Results: Between 2007 and 2020, 207 abnormal death events of physicians on the Chinese mainland were publicly reported. Among the 207 victims, the majority (~79%) died from overwork or sudden death. The number of victims who were men was 5.5 times higher than that of women, and victims were between the ages of 31–50 years. These physicians mainly belonged to the departments of surgery, anesthesiology, internal medicine, and orthopedics. Further analysis of the direct causes of death in cases of overwork death showed that 51 physicians (31.1%) died from cardiogenic diseases. Additionally, the per capita workload of physicians in China increased drastically by about 42% from 2007 to 2019, far exceeding physician workloads in Europe, Asia, and Australia (number of inpatients per physician in 2017: 72 vs. 55, 50, 45). The analysis revealed that there was a strong correlation between the number of abnormal deaths of physicians in China and the number of inpatients per physician (r = 0.683, P = 0.01).Conclusion: High-intensity working conditions may be positively correlated with the number of abnormal deaths among physicians. Smart hospital technologies have the potential to alleviate this situation.

ADDRESSING THE ELEPHANT IN THE UNDERGROUND: AN ARGUMENT FOR THE INTEGRATION OF HETEROGENEOUS DATA SOURCES FOR RECONCILIATION OF SUBSURFACE UTILITY DATA

In this paper we address the issue of unreliable subsurface utility information. Data on subsurface utilities are often positionally inaccurate, not up to date, and incomplete, leading to increased uncertainty, costs, and delays incurred in underground-related projects. Despite opportunities for improvement, the quality of legacy data remains unaddressed. We address the legacy data issue by making an argument for an approach towards subsurface utility data reconciliation that relies on the integration of heterogeneous data sources. These data sources can be collected at opportunities that occur throughout the life cycle of subsurface utilities and include as-built GIS records, GPR scans, and open excavation 3D scans. By integrating legacy data with newly captured data sources, it is possible to verify, (re)classify and update the data and improve it for future use. To demonstrate the potential of an integration-driven data reconciliation approach, we present real-world use cases from Denmark and Singapore. From these cases, challenges towards implementation of the approach were identified that include a lack of technological readiness, a lack of incentive to capture and share the data, increased cost, and data sharing concerns. Future research should investigate in detail how various data sources lead to improved data quality, develop a data model that brings together all necessary data sources for integration, and a framework for governance and master data management to ensure roles and responsibilities can be feasibly enacted.

APIs: A Common Interface for the Global Biodiversity Informatics Community

Web APIs (Application Programming Interfaces) facilitate the exchange of resources (data) between two functionally independent entities across a common programmatic interface. In more general terms, Web APIs can connect almost anything to the world wide web. Unlike traditional software, APIs are not compiled, installed, or run. Instead, data are read (or consumed in API speak) through a web-based transaction, where a client makes a request and a server responds. Web APIs can be loosely grouped into two categories within the scope of biodiversity informatics, based on purpose. First, Product APIs deliver data products to end-users. Examples include the Global Biodiversity Information Facility (GBIF) and iNaturalist APIs. Designed and built to solve specific problems, web-based Service APIs are the second type and the focus of this presentation (referred to as Service APIs). Their primary function is to provide on-demand support to existing programmatic processes. Examples of this type include Elasticsearch Suggester API and geolocation, a service that delivers geographic locations from spatial input (latitude and longitude coordinates) (Pejic et al. 2010). Many challenges lie ahead for biodiversity informatics and the sharing of global biodiversity data (e.g., Blair et al. 2020). Service-driven, standardized web-based Service APIs that adhere to best practices within the scope of biodiversity informatics can provide the transformational change needed to address many of these issues. This presentation will highlight several critical areas of interest in the biodiversity data community, describing how Service APIs can address each individually. The main topics include: standardized vocabularies, interoperability of heterogeneous data sources and data quality assessment and remediation. standardized vocabularies, interoperability of heterogeneous data sources and data quality assessment and remediation. Fundamentally, the value of any innovative technical solution can be measured by the extent of community adoption. In the context of Service APIs, adoption takes two primary forms: financial and temporal investment in the construction of clients that utilize Service APIs and willingness of the community to integrate Service APIs into their own systems and workflows. financial and temporal investment in the construction of clients that utilize Service APIs and willingness of the community to integrate Service APIs into their own systems and workflows. To achieve this, Service APIs must be simple, easy to use, pragmatic, and designed with all major stakeholder groups in mind, including users, providers, aggregators, and architects (Anderson et al. 2020Anderson et al. 2020; this study). Unfortunately, many innovative and promising technical solutions have fallen short not because of an inability to solve problems (Verner et al. 2008), rather, they were difficult to use, built in isolation, and/or designed without effective communication with stakeholders. Fortunately, projects such as Darwin Core (Wieczorek et al. 2012), the Integrated Publishing Toolkit (Robertson et al. 2014), and Megadetector (Microsoft 2021) provide the blueprint for successful community adoption of a technological solution within the biodiversity community. The final section of this presentation will examine the often overlooked non-technical aspects of this technical endeavor. Within this context, specifically how following these models can broaden community engagement and bridge the knowledge gap between the major stakeholders, resulting in the successful implementation of Service APIs.

Applying Machine Learning and Data Fusion to the "Missing Person" Problem

We present a system for integrating multiple sources of data for finding missing persons. This system can assist authorities in finding children during amber alerts, mentally challenged persons who have wandered off, or person-of-interests in an investigation. Authorities search for the person in question by reaching out to acquaintances, checking video feeds, or by looking into the previous histories relevant to the investigation. In the absence of any leads, authorities lean on public help from sources such as tweets or tip lines. A missing person investigation requires information from multiple modalities and heterogeneous data sources to be combined.<div>Existing cross-modal fusion models use separate information models for each data modality and lack the compatibility to utilize pre-existing object properties in an application domain. A framework for multimodal information retrieval, called Find-Them is developed. It includes extracting features from different modalities and mapping them into a standard schema for context-based data fusion. Find-Them can integrate application domains with previously derived object properties and can deliver data relevant for the mission objective based on the context and needs of the user. Measurements on a novel open-world cross-media dataset show the efficacy of our model. The objective of this work is to assist authorities in finding uses of Find-Them in missing person investigation.</div>

Applying Machine Learning and Data Fusion to the "Missing Person" Problem

We present a system for integrating multiple sources of data for finding missing persons. This system can assist authorities in finding children during amber alerts, mentally challenged persons who have wandered off, or person-of-interests in an investigation. Authorities search for the person in question by reaching out to acquaintances, checking video feeds, or by looking into the previous histories relevant to the investigation. In the absence of any leads, authorities lean on public help from sources such as tweets or tip lines. A missing person investigation requires information from multiple modalities and heterogeneous data sources to be combined.<div>Existing cross-modal fusion models use separate information models for each data modality and lack the compatibility to utilize pre-existing object properties in an application domain. A framework for multimodal information retrieval, called Find-Them is developed. It includes extracting features from different modalities and mapping them into a standard schema for context-based data fusion. Find-Them can integrate application domains with previously derived object properties and can deliver data relevant for the mission objective based on the context and needs of the user. Measurements on a novel open-world cross-media dataset show the efficacy of our model. The objective of this work is to assist authorities in finding uses of Find-Them in missing person investigation.</div>

Deep Learning for COVID-19 Diagnosis from CT Images

COVID-19, an infectious coronavirus disease, caused a pandemic with countless deaths. From the outset, clinical institutes have explored computed tomography as an effective and complementary screening tool alongside the reverse transcriptase-polymerase chain reaction. Deep learning techniques have shown promising results in similar medical tasks and, hence, may provide solutions to COVID-19 based on medical images of patients. We aim to contribute to the research in this field by: (i) Comparing different architectures on a public and extended reference dataset to find the most suitable; (ii) Proposing a patient-oriented investigation of the best performing networks; and (iii) Evaluating their robustness in a real-world scenario, represented by cross-dataset experiments. We exploited ten well-known convolutional neural networks on two public datasets. The results show that, on the reference dataset, the most suitable architecture is VGG19, which (i) Achieved 98.87% accuracy in the network comparison; (ii) Obtained 95.91% accuracy on the patient status classification, even though it misclassifies some patients that other networks classify correctly; and (iii) The cross-dataset experiments exhibit the limitations of deep learning approaches in a real-world scenario with 70.15% accuracy, which need further investigation to improve the robustness. Thus, VGG19 architecture showed promising performance in the classification of COVID-19 cases. Nonetheless, this architecture enables extensive improvements based on its modification, or even with preprocessing step in addition to it. Finally, the cross-dataset experiments exposed the critical weakness of classifying images from heterogeneous data sources, compatible with a real-world scenario.

BESOCIAL: A Sustainable Knowledge Graph-Based Workflow for Social Media Archiving

Social media as infrastructure for public discourse provide valuable information that needs to be preserved. Several tools for social media harvesting exist, but still only fragmented workflows may be formed with different combinations of such tools. On top of that, social media data but also preservation-related metadata standards are heterogeneous, resulting in a costly manual process. In the framework of BESOCIAL at the Royal Library of Belgium (KBR), we develop a sustainable social media archiving workflow that integrates heterogeneous data sources in a Europeana and PREMIS-based data model to describe data preserved by open source tools. This allows data stewardship on a uniform representation and we generate metadata records automatically via queries. In this paper, we present a comparison of social media harvesting tools and our Knowledge Graph-based solution which reuses off-the-shelf open source tools to harvest social media and automatically generate preservation-related metadata records. We validate our solution by generating Encoded Archival Description (EAD) and bibliographic MARC records for preservation of harvested social media collections from Twitter collected at KBR. Other archiving institutions can build upon our solution and customize it to their own social media archiving policies.

Validating deep learning inference during chest X-ray classification for COVID-19 screening

The new coronavirus unleashed a worldwide pandemic in early 2020, and a fatality rate several times that of the flu. As the number of infections soared, and capabilities for testing lagged behind, chest X-ray (CXR) imaging became more relevant in the early diagnosis and treatment planning for patients with suspected or confirmed COVID-19 infection. In a few weeks, proposed new methods for lung screening using deep learning rapidly appeared, while quality assurance discussions lagged behind. This paper proposes a set of protocols to validate deep learning algorithms, including our ROI Hide-and-Seek protocol, which emphasizes or hides key regions of interest from CXR data. Our protocol allows assessing the classification performance for anomaly detection and its correlation to radiological signatures, an important issue overlooked in several deep learning approaches proposed so far. By running a set of systematic tests over CXR representations using public image datasets, we demonstrate the weaknesses of current techniques and offer perspectives on the advantages and limitations of automated radiography analysis when using heterogeneous data sources.