Towards Structuring Clinical Texts: Joint Entity and Relation Extraction from Japanese Case Report Corpus

Important pieces of information related to patient symptoms and diagnosis are often written in free-text form in clinical texts. To utilize these texts, information extraction using natural language processing is required. This study evaluated the performance of named entity recognition (NER) and relation extraction (RE) using machine-learning methods. The Japanese case report corpus was used for this study, which had 113 types of entities and 36 types of relations that were manually annotated. There were 183 cases comprising 2,194 sentences after preprocessing. In addition, a machine learning model based on bidirectional encoder representations from transformers was used. The results revealed that the maximum micro-averaged F1 scores of NER and RE were 0.912 and 0.759, respectively. The results of this study are comparable to those of previous studies. Hence, these results could be of substantial baseline accuracy.

Digital epidemiology and infodemiology of hand-foot-mouth disease (HFMD) in Italy. Disease trend assessment via Google and Wikipedia

BACKGROUND AND AIM

The study aimed to evaluate the epidemiological trend of hand, foot and mouth disease (HFMD) in Italy using data on Internet search volume.

METHODS

A cross-sectional study design was used. Data on Internet searches were obtained from Google Trends (GT) and Wikipedia. We used the following Italian search term: "Malattia mano-piede-bocca" (Hand-foot-mouth disease, in English). A monthly time-frame was extracted, partly overlapping, from July 2015 to December 2022. GT and Wikipedia were overlapped to perform a linear regression and correlation analyses. Statistical analyses were performed using the Spearman's rank correlation coefficient (rho). A linear regression analysis was performed considering Wikipedia and GT.

RESULTS

Search peaks for both Wikipedia and GT occurred in the months November-December during the autumn-winter season and in June during the spring-summer season, except for the period from June 2020 to June 2021, probably due to the restrictions of the COVID19 pandemic. A temporal correlation was observed between GT and Wikipedia search trends.

CONCLUSIONS

This is the first study in Italy that attempts to clarify the epidemiology of HFMD. Google search and Wikipedia can be valuable for public health surveillance; however, to date, digital epidemiology cannot replace the traditional surveillance system.

Using Google Trends and Wikipedia to Investigate the Global Public's Interest in the Pancreatic Cancer Diagnosis of a Celebrity

A cross-sectional study was designed to assess the impact of a celebrity's announcement of having been diagnosed with pancreatic cancer on the volume of cancer-related research on the Internet. Global searches were carried out on Google Trends (GT) for the period from 1 January 2004 to 20 November 2022 (since data prior to 2004 were not available) using the search words Tumore del Pancreas (pancreatic cancer), Tumore neuroendocrino (neuroendocrine tumor), and Fedez (the name of a popular Italian rapper). The frequency of specific page views for Fedez, Tumore del pancreas, and Tumore neuroendocrino was collected via Wikipedia Trends data. Statistical analyses were carried out using the Pearson correlation coefficient (r). The GT data revealed a strong correlation (r = 0.83) while the Wikipedia Trends data indicated a moderate correlation (r = 0.37) for Tumore neuroendocrino and Tumore del pancreas. The search peaks for the GT and Wikipedia pages occur during the same time period. An association was found between the celebrity's announcement of his pancreatic cancer diagnosis and the volume of pancreatic-cancer-related online searches. Our findings demonstrate that media events and media coverage of health-related news can raise people's curiosity and desire for health information.

Fostering intrinsic motivation in remote undergraduate histopathology education

AIMS

The levels of abstraction, vast vocabulary and high cognitive load present significant challenges in undergraduate histopathology education. Self-determination theory describes three psychological needs which promote intrinsic motivation. This paper describes, evaluates and justifies a remotely conducted, post-COVID-19 histopathology placement designed to foster intrinsic motivation.

METHODS

90 fourth-year medical students took part in combined synchronous and asynchronous remote placements integrating virtual microscopy into complete patient narratives through Google Classroom, culminating in remote, simulated multidisciplinary team meeting sessions allowing participants to vote on 'red flag' signs and symptoms, investigations, histological diagnoses, staging and management of simulated virtual patients. The placement was designed to foster autonomy, competence and relatedness, generating authenticity, transdisciplinary integration and clinical relevance. A postpositivistic evaluation was undertaken with a validated preplacement and postplacement questionnaire capturing quantitative and qualitative data.

RESULTS

There was a significant (p<0.001) improvement in interest, confidence and competence in histopathology. Clinical integration and relevance, access to interactive resources and collaborative learning promoted engagement and sustainability post-COVID-19. Barriers to online engagement included participant lack of confidence and self-awareness in front of peers.

CONCLUSIONS

Fostering autonomy, competence and relatedness in post-COVID-19, remote educational designs can promote intrinsic motivation and authentic educational experiences. Ensuring transdisciplinary clinical integration, the appropriate use of novel technology and a focus on patient narratives can underpin the relevance of undergraduate histopathology education. The presentation of normal and diseased tissue in this way can serve as an important mode for the acquisition and application of clinically relevant knowledge expected of graduates.

Applications of machine learning in the chemical pathology laboratory

Machine learning (ML) is an area of artificial intelligence that provides computer programmes with the capacity to autodidact and learn new skills from experience, without continued human programming. ML algorithms can analyse large data sets quickly and accurately, by supervised and unsupervised learning techniques, to provide classification and prediction value outputs. The application of ML to chemical pathology can potentially enhance efficiency at all phases of the laboratory's total testing process. Our review will broadly discuss the theoretical foundation of ML in laboratory medicine. Furthermore, we will explore the current applications of ML to diverse chemical pathology laboratory processes, for example, clinical decision support, error detection in the preanalytical phase, and ML applications in gel-based image analysis and biomarker discovery. ML currently demonstrates exploratory applications in chemical pathology with promising advancements, which have the potential to improve all phases of the chemical pathology total testing pathway.

New life for old cellular pathology: a transformational approach to the upcycling of historic e-pathology records for contemporary clinical uses

Aims

Cellular pathology (‘e-pathology’) record sets are a rich data resource with which to populate the electronic patient record (EPR). Accessible reports, even decades old, can be of great value in contemporary clinical decision making and as a resource for longitudinal clinical research. The aim of this short paper is to describe a solution in a major UK University Hospital which gives immediate visibility and clinical utility to 30 years of e-pathology records

Methods

Over the past decade, we have created a timeline structured and iconographic data framework for the ‘whole-of-life’ visualisation of the entirety of an EPR. We have enhanced this interface with the sequential extraction of 373 342 e-pathology reports from legacy Ferranti (1990–1997) and Masterlab (1997–2004) files. They have been uploaded into our SQL file servers, following appropriate data quality and patient identity reconciliation checks.

Results

We have restored a large repository of previously inaccessible e-pathology records to clinical use and to immediacy of access as a foundation element of our timeline structured EPR. This process has also allowed us to populate and validate an EPR-integral breast cancer data system of 20 000 cases with e-pathology records dating back to 1990.

Conclusions

The revitalisation of old e-pathology reports into a timeline structured EPR creates preserves and upcycles the investment in pathology reporting which is otherwise progressively lost to clinical use. E-pathology records provide reliable, life-long evidence of critical transition points in individual lives and disease progression for clinical and research use, when they can be instantly accessed.

[Potential for Big Data Analysis Using AI in the Field of Clinical Pharmacy]

Industrial reforms utilizing artificial intelligence (AI) have advanced remarkably in recent years. The application of AI to big data analysis in the medical information field has also been advancing and is expected to be used to find drug adverse effects that cannot be predicted by conventional methods. We have developed an adverse drug reactions analysis system that uses machine learning and data from the Japanese Adverse Drug Event Report (JADER) database. The system was developed using the C# programming language and incorporates the open source machine learning library Accord.Net. Potential analytical capabilities of the system include discovering unknown drug adverse effects and evaluating drug-induced adverse events in pharmaceutical management. However, to apply the system to pharmaceutical management, it is important to examine the characteristics and suitability of the level of AI used in the system and to select statistical methods or machine learning when appropriate. If these points are addressed, there is potential for pharmaceutical management to be individualized and optimized in the clinical setting by using the developed system to analyze big data. The system also has the potential to allow individual healthcare facilities such as hospitals and pharmacies to contribute to drug repositioning, including the discovery of new efficacies, interactions, and drug adverse events.

A reference guide to rapidly implementing an institutional dashboard for resource allocation and oversight during COVID-19 pandemic surge

Objectives

We develop a dashboard that leverages electronic health record (EHR) data to monitor intensive care unit patient status and ventilator utilization in the setting of the COVID-19 pandemic.

Materials and methods

Data visualization software is used to display information from critical care data mart that extracts information from the EHR. A multidisciplinary collaborative led the development.

Results

The dashboard displays institution-level ventilator utilization details, as well as patient-level details such as ventilator settings, organ-system specific parameters, laboratory values, and infusions.

Discussion

Components of the dashboard were selected to facilitate the determination of resources and simultaneous assessment of multiple patients. Abnormal values are color coded. An overall illness assessment score is tracked daily to capture illness severity over time.

Conclusion

This reference guide shares the architecture and sample reusable code to implement a robust, flexible, and scalable dashboard for monitoring ventilator utilization and illness severity in intensive care unit ventilated patients.

Initial study of anaemia profile for primary care centres with automated laboratory algorithms reduces the demand for ferritin, iron, transferrin, vitamin B12 and folate tests

AIM

To evaluate the influence of an algorithm designed to incorporate reflex testing according to haemogram results for analytical tests ordered to investigate anaemia.

METHODS

In 2020, a new request for 'initial study of anaemia' was created in three primary care pilot centres for suspected anaemia or new anaemias. A haemogram was ordered and the remainder of the tests were created in a reflex manner according to an algorithm integrated in the laboratory information system that also generates a comment that is completed and validated by a haematologist. The demand for tests was evaluated over three time periods.

RESULTS

Of 396 requests, anaemia was detected in 80 (20.2%), with 26 microcytic anaemias (6.57%), 20 iron deficiency anaemias, 41 (10.3%) normocytic anaemias and 13 macrocytic anaemias (3.28%); 4 with folate deficiency; and 1 haemolytic anaemia. No haematological diseases were detected. Twenty-four (6.06%) cases exhibited microcytosis/hypochromia without anaemia, 12 of which exhibited iron deficiency. Four young women exhibiting within-limit haemoglobin levels had iron deficiency. There were 56 (14.1%) cases of macrocytosis without anaemia.With the new profile of 'initial study of anaemia', the demand for tests was reduced and was significantly lower than in the remainder of primary centres for iron, transferrin, ferritin, vitamin B₁₂ and folate.

CONCLUSIONS

A new profile of 'initial study of anaemia' in the request form with algorithms integrated in the laboratory information system enabled submission of orders and decreased the demand for unnecessary iron, transferrin, ferritin, vitamin B₁₂ and folate tests.

Rocky road to digital diagnostics: implementation issues and exhilarating experiences

Since 2007, we have gradually been building up infrastructure for digital pathology, starting with a whole slide scanner park to build up a digital archive to streamline doing multidisciplinary meetings, student teaching and research, culminating in a full digital diagnostic workflow where we are currently integrating artificial intelligence algorithms. In this paper, we highlight the different steps in this process towards digital diagnostics, which was at times a rocky road with definitely issues in implementation, but eventually an exciting new way to practice pathology in a more modern and efficient way where patient safety has clearly gone up.

Digital pathology in the time of corona

The 2020 COVID-19 crisis has had and will have many implications for healthcare, including pathology. Rising number of infections create staffing shortages and other hospital departments might require pathology employees to fill more urgent positions. Furthermore, lockdown measures and social distancing cause many people to work from home. During this crisis, it became clearer than ever what an asset digital diagnostics is to keep pathologists, residents, molecular biologists and pathology assistants engaged in the diagnostic process, allowing social distancing and a ‘need to be there’ on-the-premises policy, while working effectively from home. This paper provides an overview of our way of working during the 2020 COVID-19 crisis with emphasis on the virtues of digital pathology.

Engagement and Participant Experiences With Consumer Smartwatches for Health Research: Longitudinal, Observational Feasibility Study

BACKGROUND

Wearables provide opportunities for frequent health data collection and symptom monitoring. The feasibility of using consumer cellular smartwatches to provide information both on symptoms and contemporary sensor data has not yet been investigated.

OBJECTIVE

This study aimed to investigate the feasibility and acceptability of using cellular smartwatches to capture multiple patient-reported outcomes per day alongside continuous physical activity data over a 3-month period in people living with knee osteoarthritis (OA).

METHODS

For the KOALAP (Knee OsteoArthritis: Linking Activity and Pain) study, a novel cellular smartwatch app for health data collection was developed. Participants (age ≥50 years; self-diagnosed knee OA) received a smartwatch (Huawei Watch 2) with the KOALAP app. When worn, the watch collected sensor data and prompted participants to self-report outcomes multiple times per day. Participants were invited for a baseline and follow-up interview to discuss their motivations and experiences. Engagement with the watch was measured using daily watch wear time and the percentage completion of watch questions. Interview transcripts were analyzed using grounded thematic analysis.

RESULTS

A total of 26 people participated in the study. Good use and engagement were observed over 3 months: most participants wore the watch on 75% (68/90) of days or more, for a median of 11 hours. The number of active participants declined over the study duration, especially in the final week. Among participants who remained active, neither watch time nor question completion percentage declined over time. Participants were mainly motivated to learn about their symptoms and enjoyed the self-tracking aspects of the watch. Barriers to full engagement were battery life limitations, technical problems, and unfulfilled expectations of the watch. Participants reported that they would have liked to report symptoms more than 4 or 5 times per day.

CONCLUSIONS

This study shows that capture of patient-reported outcomes multiple times per day with linked sensor data from a smartwatch is feasible over at least a 3-month period.

INTERNATIONAL REGISTERED REPORT IDENTIFIER (IRRID)

RR2-10.2196/10238.

Computational Fluid Dynamic Modeling of Urethral Strictures

PURPOSE

Computational fluid dynamics have paradigm shifting potential in understanding the physiological flow of fluids in the human body. This translational branch of engineering has already made an important clinical impact on the study of cardiovascular disease. We evaluated the feasibility and applicability of computational fluid dynamics to model urine flow.

MATERIALS AND METHODS

We prepared a computational fluid dynamics model using an idealized male genitourinary system. We created 16 hypothetical urethral stricture scenarios as a test bed. Standard parameters of urine such as pressure, temperature and viscosity were applied as well as typical assumptions germane to fluid dynamic modeling. We used ABAQUS/CAE 6.14 (Dassault Systèmes®) with a direct unsymmetrical solver with standard (FC3D8) 3D brick 8Node elements for model generation.

RESULTS

The average flow rate in urethral stricture disease as measured by our model was 5.97 ml per second (IQR 2.2-10.9). The model predicted a flow rate of 2.88 ml per second for a single 5Fr stricture in the mid bulbar urethra when assuming all other variables constant. The model demonstrated that increasing stricture diameter and bladder pressure strongly impacted urine flow while stricture location and length, and the sequence of multiple strictures had a weaker impact.

CONCLUSIONS

We successfully created a computational fluid dynamics model of an idealized male urethra with varied types of urethral strictures. The resultant flow rates were consistent with the literature. The accuracy of modeling increasing bladder pressure should be improved by future iterations. This technology has vast research and clinical potential.

Health Care and Precision Medicine Research: Analysis of a Scalable Data Science Platform

Background

Health care data are increasing in volume and complexity. Storing and analyzing these data to implement precision medicine initiatives and data-driven research has exceeded the capabilities of traditional computer systems. Modern big data platforms must be adapted to the specific demands of health care and designed for scalability and growth.

Objective

The objectives of our study were to (1) demonstrate the implementation of a data science platform built on open source technology within a large, academic health care system and (2) describe 2 computational health care applications built on such a platform.

Methods

We deployed a data science platform based on several open source technologies to support real-time, big data workloads. We developed data-acquisition workflows for Apache Storm and NiFi in Java and Python to capture patient monitoring and laboratory data for downstream analytics.

Results

Emerging data management approaches, along with open source technologies such as Hadoop, can be used to create integrated data lakes to store large, real-time datasets. This infrastructure also provides a robust analytics platform where health care and biomedical research data can be analyzed in near real time for precision medicine and computational health care use cases.

Conclusions

The implementation and use of integrated data science platforms offer organizations the opportunity to combine traditional datasets, including data from the electronic health record, with emerging big data sources, such as continuous patient monitoring and real-time laboratory results. These platforms can enable cost-effective and scalable analytics for the information that will be key to the delivery of precision medicine initiatives. Organizations that can take advantage of the technical advances found in data science platforms will have the opportunity to provide comprehensive access to health care data for computational health care and precision medicine research.

Collecting Symptoms and Sensor Data With Consumer Smartwatches (the Knee OsteoArthritis, Linking Activity and Pain Study): Protocol for a Longitudinal, Observational Feasibility Study

BACKGROUND

The Knee OsteoArthritis, Linking Activity and Pain (KOALAP) study is the first to test the feasibility of using consumer-grade cellular smartwatches for health care research.

OBJECTIVE

The overall aim was to investigate the feasibility of using consumer-grade cellular smartwatches as a novel tool to capture data on pain (multiple times a day) and physical activity (continuously) in patients with knee osteoarthritis. Additionally, KOALAP aimed to investigate smartwatch sensor data quality and assess whether engagement, acceptability, and user experience are sufficient for future large-scale observational and interventional studies.

METHODS

A total of 26 participants with self-diagnosed knee osteoarthritis were recruited in September 2017. All participants were aged 50 years or over and either lived in or were willing to travel to the Greater Manchester area. Participants received a smartwatch (Huawei Watch 2) with a bespoke app that collected patient-reported outcomes via questionnaires and continuous watch sensor data. All data were collected daily for 90 days. Additional data were collected through interviews (at baseline and follow-up) and baseline and end-of-study questionnaires. This study underwent full review by the University of Manchester Research Ethics Committee (#0165) and University Information Governance (#IGRR000060). For qualitative data analysis, a system-level security policy was developed in collaboration with the University Information Governance Office. Additionally, the project underwent an internal review process at Google, including separate reviews of accessibility, product engineering, privacy, security, legal, and protection regulation compliance.

RESULTS

Participants were recruited in September 2017. Data collection via the watches was completed in January 2018. Collection of qualitative data through patient interviews is still ongoing. Data analysis will commence when all data are collected; results are expected in 2019.

CONCLUSIONS

KOALAP is the first health study to use consumer cellular smartwatches to collect self-reported symptoms alongside sensor data for musculoskeletal disorders. The results of this study will be used to inform the design of future mobile health studies. Results for feasibility and participant motivations will inform future researchers whether or under which conditions cellular smartwatches are a useful tool to collect patient-reported outcomes alongside passively measured patient behavior. The exploration of associations between self-reported symptoms at different moments will contribute to our understanding of whether it may be valuable to collect symptom data more frequently. Sensor data-quality measurements will indicate whether cellular smartwatch usage is feasible for obtaining sensor data. Methods for data-quality assessment and data-processing methods may be reusable, although generalizability to other clinical areas should be further investigated.

INTERNATIONAL REGISTERED REPORT IDENTIFIER (IRRID)

DERR1-10.2196/10238.

Triaging patients at risk of influenza using a patient portal

Vanderbilt University has a widely adopted patient portal, MyHealthAtVanderbilt, which provides an infrastructure to deliver information that can empower patient decision making and enhance personalized healthcare. An interdisciplinary team has developed Flu Tool, a decision-support application targeted to patients with influenza-like illness and designed to be integrated into a patient portal. Flu Tool enables patients to make informed decisions about the level of care they require and guides them to seek timely treatment as appropriate. A pilot version of Flu Tool was deployed for a 9-week period during the 2010-2011 influenza season. During this time, Flu Tool was accessed 4040 times, and 1017 individual patients seen in the institution were diagnosed as having influenza. This early experience with Flu Tool suggests that healthcare consumers are willing to use patient-targeted decision support. The design, implementation, and lessons learned from the pilot release of Flu Tool are described as guidance for institutions implementing decision support through a patient portal infrastructure.

MyHealthAtVanderbilt: policies and procedures governing patient portal functionality

Explicit guidelines are needed to develop safe and effective patient portals. This paper proposes general principles, policies, and procedures for patient portal functionality based on MyHealthAtVanderbilt (MHAV), a robust portal for Vanderbilt University Medical Center. We describe policies and procedures designed to govern popular portal functions, address common user concerns, and support adoption. We present the results of our approach as overall and function-specific usage data. Five years after implementation, MHAV has over 129,800 users; 45% have used bi-directional messaging; 52% have viewed test results and 45% have viewed other medical record data; 30% have accessed health education materials; 39% have scheduled appointments; and 29% have managed a medical bill. Our policies and procedures have supported widespread adoption and use of MHAV. We believe other healthcare organizations could employ our general guidelines and lessons learned to facilitate portal implementation and usage.

A framework for comparing phenotype annotations of orthologous genes

OBJECTIVES

Animal models are a key resource for the investigation of human diseases. In contrast to functional annotation, phenotype annotation is less standard, and comparing phenotypes across species remains challenging. The objective of this paper is to propose a framework for comparing phenotype annotations of orthologous genes based on the Medical Subject Headings (MeSH) indexing of biomedical articles in which these genes are discussed.

METHODS

17,769 pairs of orthologous genes (mouse and human) are downloaded from the Mouse Genome Informatics (MGI) system and linked to biomedical articles through Entrez Gene. MeSH index terms corresponding to diseases are extracted from Medline.

RESULTS

11,111 pairs of genes exhibited at least one phenotype annotation for each gene in the pair. Among these, 81% have at least one phenotype annotation in common, 80% have at least one annotation specific to the human gene and 84% have at least one annotation specific to the mouse gene. Four disease categories represent 54% of all phenotype annotations.

CONCLUSIONS

This framework supports the curation of phenotype annotation and the generation of research hypotheses based on comparative studies.