A guide to navigating administrative data linkage for research.

Data linkage brings together information from various sources, including routinely collected administrative data or data from different research studies, to create a new, richer dataset. It provides insights into complex relationships between health and outcomes and evidence pathways to good health. However, when considering data linkage, there are several processes and practicality aspects that need to be explored. Some of these include understanding the costs, complexity of linkage, data storage requirements, required applications, and time lags. Taking these practicalities into consideration will lead to a more efficient process for data linkage.

Data drift in medical machine learning: implications and potential remedies.

Data drift refers to differences between the data used in training a machine learning (ML) model and that applied to the model in real-world operation. Medical ML systems can be exposed to various forms of data drift, including differences between the data sampled for training and used in clinical operation, differences between medical practices or context of use between training and clinical use, and time-related changes in patient populations, disease patterns, and data acquisition, to name a few. In this article, we first review the terminology used in ML literature related to data drift, define distinct types of drift, and discuss in detail potential causes within the context of medical applications with an emphasis on medical imaging. We then review the recent literature regarding the effects of data drift on medical ML systems, which overwhelmingly show that data drift can be a major cause for performance deterioration. We then discuss methods for monitoring data drift and mitigating its effects with an emphasis on pre- and post-deployment techniques. Some of the potential methods for drift detection and issues around model retraining when drift is detected are included. Based on our review, we find that data drift is a major concern in medical ML deployment and that more research is needed so that ML models can identify drift early, incorporate effective mitigation strategies and resist performance decay.

Lessons learned from urgent computing in Europe: Tackling the COVID-19 pandemic.

PRACE (Partnership for Advanced Computing in Europe), an international not-for-profit association that brings together the five largest European supercomputing centers and involves 26 European countries, has allocated more than half a billion core hours to computer simulations to fight the COVID-19 pandemic. Alongside experiments, these simulations are a pillar of research to assess the risks of different scenarios and investigate mitigation strategies. While the world deals with the subsequent waves of the pandemic, we present a reflection on the use of urgent supercomputing for global societal challenges and crisis management.

Incorporating medical code descriptions for diagnosis prediction in healthcare.

BACKGROUND: Diagnosis aims to predict the future health status of patients according to their historical electronic health records (EHR), which is an important yet challenging task in healthcare informatics. Existing diagnosis prediction approaches mainly employ recurrent neural networks (RNN) with attention mechanisms to make predictions. However, these approaches ignore the importance of code descriptions, i.e., the medical definitions of diagnosis codes. We believe that taking diagnosis code descriptions into account can help the state-of-the-art models not only to learn meaning code representations, but also to improve the predictive performance, especially when the EHR data are insufficient. METHODS: We propose a simple, but general diagnosis prediction framework, which includes two basic components: diagnosis code embedding and predictive model. To learn the interpretable code embeddings, we apply convolutional neural networks (CNN) to model medical descriptions of diagnosis codes extracted from online medical websites. The learned medical embedding matrix is used to embed the input visits into vector representations, which are fed into the predictive models. Any existing diagnosis prediction approach (referred to as the base model) can be cast into the proposed framework as the predictive model (called the enhanced model). RESULTS: We conduct experiments on two real medical datasets: the MIMIC-III dataset and the Heart Failure claim dataset. Experimental results show that the enhanced diagnosis prediction approaches significantly improve the prediction performance. Moreover, we validate the effectiveness of the proposed framework with insufficient EHR data. Finally, we visualize the learned medical code embeddings to show the interpretability of the proposed framework. CONCLUSIONS: Given the historical visit records of a patient, the proposed framework is able to predict the next visit information by incorporating medical code descriptions.

[Willingness to Acceptance for the Information Leakage of Medical Information Data Using the Contingent Valuation Method].

Appropriate information security measures are very important for today's highly computerized hospitals to maintain the trust from patients. If once the personal information leakage of medical information was occurred, the hospital could lose their trust that has built for long time so far. It is important for hospitals to know the impact of the leakage accident previously advance to decide the investment for information security. The purpose of this study is to evaluate the impact of medical information leakage. The comforting fee for the patient's mental damage as the willingness to accept (WTA) was estimated, when the information leak occurred from a hospital using the contingent valuation method (CVM). Questionnaire survey was conducted using an internet survey service in Japan. We asked for 300 citizens about the use of personal information communication equipment and information security measures and their awareness for the information leakage. In addition, we presented a hypothetical scenario regarding information leakage of own medical information, asked the WTA as the comforting fee by the one choice of acceptance or rejection for the presented fee. In 300 responses, 190 were could be used for WTA estimation. WTA as the comforting fee when the information leakage of medical care information occurred, was estimated 570,541 yen in total. The result was similar with the value estimated by the damage compensation payment estimation model.

Obtención automática de palabras clave en textos clínicos: una aplicación de procesamiento del lenguaje natural a datos masivos de sospecha diagnóstica en Chile / Automatic keyword retrieval from clinical texts: an application of natural language processing to massive data of Chilean suspected diagnosis

Background: Free-text imposes a challenge in health data analysis since the lack of structure makes the extraction and integration of information difficult, particularly in the case of massive data. An appropriate machine-interpretation of electronic health records in Chile can unleash knowledge contained in large volumes of clinical texts, expanding clinical management and national research capabilities. Aim: To illustrate the use of a weighted frequency algorithm to find keywords. This finding was carried out in the diagnostic suspicion field of the Chilean specialty consultation waiting list, for diseases not covered by the Chilean Explicit Health Guarantees plan. Material and Methods: The waiting lists for a first specialty consultation for the period 2008-2018 were obtained from 17 out of 29 Chilean health services, and total of 2,592,925 diagnostic suspicions were identified. A natural language processing technique called Term Frequency-Inverse Document Frequency was used for the retrieval of diagnostic suspicion keywords. Results: For each specialty, four key words with the highest weighted frequency were determined. Word clouds showing words weighted by their importance were created to obtain a visual representation. These are available at cimt.uchile.cl/lechile/. Conclusions: The algorithm allowed to summarize unstructured clinical free-text data, improving its usefulness and accessibility.

Convolution kernel and iterative reconstruction affect the diagnostic performance of radiomics and deep learning in lung adenocarcinoma pathological subtypes.

BACKGROUND: The aim of this study was to investigate the influence of convolution kernel and iterative reconstruction on the diagnostic performance of radiomics and deep learning (DL) in lung adenocarcinomas. METHODS: A total of 183 patients with 215 lung adenocarcinomas were included in this study. All CT imaging data was reconstructed with three reconstruction algorithms (ASiR at 0%, 30%, 60% strength), each with two convolution kernels (bone and standard). A total of 171 nodules were selected as the training-validation set, whereas 44 nodules were selected as the testing set. Logistic regression and a DL framework-DenseNets were selected to tackle the task. Three logical experiments were implemented to fully explore the influence of the studied parameters on the diagnostic performance. The receiver operating characteristic curve (ROC) was used to evaluate the performance of constructed models. RESULTS: In Experiments A and B, no statistically significant results were found in the radiomic method, whereas two and six pairs were statistically significant (P < 0.05) in the DL method. In Experiment_C, significant differences in one and four models were found in the radiomics and DL methods, respectively. Moreover, models constructed with standard convolution kernel data outperformed that constructed with bone convolution kernel data in all studied ASiR levels in the DL method. In the DL method, B0 and S60 performed best in bone and standard convolution kernel, respectively. CONCLUSION: The results demonstrated that DL was more susceptible to CT parameter variability than radiomics. Standard convolution kernel images seem to be more appropriate for imaging analysis. Further investigation with a larger sample size is needed.

[The importance of digitalisation in urology]. / Stellenwert der Digitalisierung im urologischen Alltag.

The Internet has shaped and changed society like no other technology. Culturally, the emergence of the Internet is being described as having the same impact on society as the invention of printing. In 2018, more than 4 billion people had access to the Internet. Among all Internet users, approximately 80 % search the Internet for health-related information, with cancer being the most frequently searched condition. Patients rate the Internet as the second most helpful source of information, outranked only by consultation with a medical doctor. There are more than 2.6 billion active social media users. Among urological residents, 97 % use social media on a regular basis. Digitalisation has the potential to strengthen patients' health literacy and optimise patient care, especially in the oncologic field. In summary, digitalisation bears an enormous potential for the field of urology.

Removing Confounding Factors Associated Weights in Deep Neural Networks Improves the Prediction Accuracy for Healthcare Applications.

The proliferation of healthcare data has brought the opportunities of applying data-driven approaches, such as machine learning methods, to assist diagnosis. Recently, many deep learning methods have been shown with impressive successes in predicting disease status with raw input data. However, the "black-box" nature of deep learning and the highreliability requirement of biomedical applications have created new challenges regarding the existence of confounding factors. In this paper, with a brief argument that inappropriate handling of confounding factors will lead to models' sub-optimal performance in real-world applications, we present an efficient method that can remove the inuences of confounding factors such as age or gender to improve the across-cohort prediction accuracy of neural networks. One distinct advantage of our method is that it only requires minimal changes of the baseline model's architecture so that it can be plugged into most of the existing neural networks. We conduct experiments across CT-scan, MRA, and EEG brain wave with convolutional neural networks and LSTM to verify the efficiency of our method.

Process Mining Dashboard in Operating Rooms: Analysis of Staff Expectations with Analytic Hierarchy Process.

The widespread adoption of real-time location systems is boosting the development of software applications to track persons and assets in hospitals. Among the vast amount of applications, real-time location systems in operating rooms have the advantage of grounding advanced data analysis techniques to improve surgical processes, such as process mining. However, such applications still find entrance barriers in the clinical context. In this paper, we aim to evaluate the preferred features of a process mining-based dashboard deployed in the operating rooms of a hospital equipped with a real-time location system. The dashboard allows to discover and enhance flows of patients based on the location data of patients undergoing an intervention. Analytic hierarchy process was applied to quantify the prioritization of the dashboard features (filtering data, enhancement, node selection, statistics, etc.), distinguishing the priorities that each of the different roles in the operating room service assigned to each feature. The staff in the operating rooms (n = 10) was classified into three groups: Technical, clinical, and managerial staff according to their responsibilities. Results showed different weights for the features in the process mining dashboard for each group, suggesting that a flexible process mining dashboard is needed to boost its potential in the management of clinical interventions in operating rooms. This paper is an extension of a communication presented in the Process-Oriented Data Science for Health Workshop in the Business Process Management Conference 2018.

[Automatic keyword retrieval from clinical texts: an application of natural language processing to massive data of Chilean suspected diagnosis]. / Obtención automática de palabras clave en textos clínicos: una aplicación de procesamiento del lenguaje natural a datos masivos de sospecha diagnóstica en Chile.

BACKGROUND: Free-text imposes a challenge in health data analysis since the lack of structure makes the extraction and integration of information difficult, particularly in the case of massive data. An appropriate machine-interpretation of electronic health records in Chile can unleash knowledge contained in large volumes of clinical texts, expanding clinical management and national research capabilities. AIM: To illustrate the use of a weighted frequency algorithm to find keywords. This finding was carried out in the diagnostic suspicion field of the Chilean specialty consultation waiting list, for diseases not covered by the Chilean Explicit Health Guarantees plan. MATERIAL AND METHODS: The waiting lists for a first specialty consultation for the period 2008-2018 were obtained from 17 out of 29 Chilean health services, and total of 2,592,925 diagnostic suspicions were identified. A natural language processing technique called Term Frequency-Inverse Document Frequency was used for the retrieval of diagnostic suspicion keywords. RESULTS: For each specialty, four key words with the highest weighted frequency were determined. Word clouds showing words weighted by their importance were created to obtain a visual representation. These are available at cimt.uchile.cl/lechile/. CONCLUSIONS: The algorithm allowed to summarize unstructured clinical free-text data, improving its usefulness and accessibility.

PACS Implementation Challenges in a Public Healthcare Institution: A South African Vendor Perspective / 대한의료정보학회지

OBJECTIVES: Conventional radiological processes have been replaced by digital images and information technology systems within South Africa and other developing countries. Picture Archiving and Communication Systems (PACS) technology offers many benefits to institutions, medical personnel and patients; however, the implementation of such systems can be a challenging task. It has been documented that South Africa has been using PACS for more than a decade in public hospitals with moderate success. The aim of this study was to identify and describe the PACS challenges endured by PACS vendors during implementation in the South African public healthcare sector. METHODS: This was achieved by engaging in a methodological approach that was qualitative in nature collecting data through semi structured interviews from 10 PACS experts/participants which were later analysed qualitatively. RESULTS: The findings show that PACS vendors have countless challenges, some of which include space, insufficient infrastructure, image storage capacity, system maturity and vendor related concerns. It was clear that the PACS experts readily offered contextually appropriate descriptions of their encounters during PACS implementations in South African public healthcare institutions. CONCLUSIONS: PACS vendors anticipate these challenges when facing a public healthcare institution and it is recommended that the hospital management and potential PACS stakeholders be made aware of these challenges to mitigate their effects and aid in a successful implementation.

Early-warning of ARDS using novelty detection and data fusion.

Acute respiratory distress syndrome (ARDS) is a critical condition that disturbs the respiratory system and may lead to death. Early identification of this syndrome is crucial for the implementation of preventive measures. The present paper focuses on the prediction of the onset of this syndrome using physiological records of patients. Heart rate, respiratory rate, peripheral arterial oxygen saturation and mean airway blood pressure were considered. The method proposed in this paper uses first distance-based novelty detection that allows detecting deviations from normal states for each signal. Then, linear and nonlinear kernel-based data fusion algorithms are introduced to combine the individual signal decisions. The proposed method is evaluated using the MIMIC II physiological database. As a result, ARDS is detected in the early phases of occurrence with sensitivity and specificity of 65% and 100% respectively for the combination of all the signals in study. Moreover, the proposed method outperforms current state-of-the-art methods in real-time surveillance of ARDS using only physiological data with an average prediction before 39 h of onset.

Implementation of informatics for integrating biology and the bedside (i2b2) platform as Docker containers.

BACKGROUND: Informatics for Integrating Biology and the Bedside (i2b2) is an open source clinical data analytics platform used at over 200 healthcare institutions for querying patient data. The i2b2 platform has several components with numerous dependencies and configuration parameters, which renders the task of installing or upgrading i2b2 a challenging one. Even with the availability of extensive documentation and tutorials, new users often require several weeks to correctly install a functional i2b2 platform. The goal of this work is to simplify the installation and upgrade process for i2b2. Specifically, we have containerized the core components of the platform, and evaluated the containers for ease of installation. RESULTS: We developed three Docker container images: WildFly, database, and web, to encapsulate the three major deployment components of i2b2. These containers isolate the core functionalities of the i2b2 platform, and work in unison to provide its functionalities. Our evaluations indicate that i2b2 containers function successfully on the Linux platform. Our results demonstrate that the containerized components work out-of-the-box, with minimal configuration. CONCLUSIONS: Containerization offers the potential to package the i2b2 platform components into standalone executable packages that are agnostic to the underlying host operating system. By releasing i2b2 as a Docker container, we anticipate that users will be able to create a working i2b2 hive installation without the need to download, compile, and configure individual components that constitute the i2b2 cells, thus making this platform accessible to a greater number of institutions.

Protecting Privacy and Security of Genomic Data in i2b2 with Homomorphic Encryption and Differential Privacy.

Re-use of patients' health records can provide tremendous benefits for clinical research. Yet, when researchers need to access sensitive/identifying data, such as genomic data, in order to compile cohorts of well-characterized patients for specific studies, privacy and security concerns represent major obstacles that make such a procedure extremely difficult if not impossible. In this paper, we address the challenge of designing and deploying in a real operational setting an efficient privacy-preserving explorer for genetic cohorts. Our solution is built on top of the i2b2 (Informatics for Integrating Biology and the Bedside) framework and leverages cutting-edge privacy-enhancing technologies such as homomorphic encryption and differential privacy. Solutions involving homomorphic encryption are often believed to be costly and immature for use in operational environments. Here, we show that, for specific applications, homomorphic encryption is actually a very efficient enabler. Indeed, our solution outperforms prior work by enabling a researcher to securely compute simple statistics on more than 3,000 encrypted genetic variants simultaneously for a cohort of 5,000 individuals in less than 5 seconds with commodity hardware. To the best of our knowledge, our privacy-preserving solution is the first to also be successfully deployed and tested in a operation setting (Lausanne University Hospital).

Informative graphing of continuous safety variables relative to normal reference limits.

BACKGROUND: Interpreting graphs of continuous safety variables can be complicated because differences in age, gender, and testing site methodologies data may give rise to multiple reference limits. Furthermore, data below the lower limit of normal are compressed relative to those points above the upper limit of normal. The objective of this study is to develop a graphing technique that addresses these issues and is visually intuitive. METHODS: A mock dataset with multiple reference ranges is initially used to develop the graphing technique. Formulas are developed for conditions where data are above the upper limit of normal, normal, below the lower limit of normal, and below the lower limit of normal when the data value equals zero. After the formulae are developed, an anonymized dataset from an actual set of trials for an approved drug is evaluated comparing the technique developed in this study to standard graphical methods. RESULTS: Formulas are derived for the novel graphing method based on multiples of the normal limits. The formula for values scaled between the upper and lower limits of normal is a novel application of a readily available scaling formula. The formula for the lower limit of normal is novel and addresses the issue of this value potentially being indeterminate when the result to be scaled as a multiple is zero. CONCLUSIONS: The formulae and graphing method described in this study provides a visually intuitive method to graph continuous safety data including laboratory values, vital sign data.

[Homo digitalis and occupational health]. / Homo numericus et santé au travail.

Although the digital transformation is widely viewed as a lever for productivity, it is now also seen, in public occupational health policies, as a factor for the potential increase in occupational risks. This digital transformation can also provide possible solutions enabling the homo digitalis to avoid these risks.