Programme

Chair: Sumithra Velupillai

1. Mark Ormerod, Jesús Martínez del Rincón, Neil Robertson, Bernadette McGuinness, Barry Devereux: Assessing the interpretability of sentence-level clinical notes diagnosis models (slides)
2. Julia Walsh, Jonathan Cave, Frances Griffiths: Using spontaneously generated online patient experiences to improve healthcare for patients and providers (slides)
3. Glorianna Jagfeld, Steven Jones, Fiona Lobban, Paul Rayson: Talking about recovery in bipolar disorder (slides)

Panelists: Nigel Collier (University of Cambridge), Hongfang Liu (Mayo Clinic), Stephane Meystre (Medical University of South Carolina)

1. Philip John Gorinski, Honghan Wu, Claire Grover, Richard Tobin, Conn Talbot, Heather Whalley, Cathie Sudlow, William Whiteley and Beatrice Alex. Named entity recognition for EHRs: A comparison of rule-based and machine learning approaches (slides)
2. Noa Cruz, Sergio Collazo, Ana López-Ballesteros and Ignacio Hernández-Medrano. Annotation of atherosclerotic/cardiovascular clinical entities from EHRs (slides)
3. Antoine Pironet, Joris Mattheijssens, Kris Henau, Nancy Van Damme, Hélène Antoine-Poirel, Harlinde de Schutter and Liesbet Van Eycken. Automatic extraction of breast receptor status from bilingual free-text cancer pathology reports (slides)
4. Amal Alharbi and Mark Stevenson. Using query adaptation to improve the identification of relevant studies for systematic reviews (slides)

Insufficient patient enrollment in clinical trials remains a serious and costly problem and is often considered the most critical issue to solve for the clinical research community. Most patients are never offered an opportunity to enroll in clinical trials, causing wasted time, resources and potential survival extension and quality of life improvements. The participation of physicians is essential for successful patient enrollment and lack of awareness of trials is often cited as a reason for low enrollment levels. One potential barrier to this awareness is the difficulty in correlating eligibility criteria with patient characteristics in a timely manner. Automated eligibility criteria extraction from narrative trial descriptions and extraction of matching clinical information from patient electronic health records (EHR) have been explored to address the aforementioned issues, but both approaches suffer from limited scope and generalizability. More importantly, no fully automated matching between trials and patients for trial eligibility have been attempted. In this talk, I will present our vision to improve patient enrollment in clinical trials and some foundational work to automatically extract eligibility criteria from EHR text notes and to match trials with eligible patients automatically. (slides)

Elizabeth Ford, Kerina Jones, Lamiece Hassan, Anoop Shah, Nathan Lea

One of the biggest barriers to progressing the field of healthcare text analytics is gaining access to patient free text data from medical records, reports and letters. Data custodians may be wary of allowing this data to be shared for research because of increased privacy risks and because no evidence exists about public opinions of medical text data being shared for research. In addition, where researchers aim to process text relating to health from sources such as social media and online discussion forums, there is still some confusion about which ethics principles and approval processes to follow. In this panel session we will present current work within the Healtex community on four projects: The Brighton citizens’ jury commissioned by Healtex in June 2018; the ongoing TexGov project which is working with the research community and the public to develop data governance standards for using free text; a summary of ethical commentaries and public views on scraping social media and internet data to use for health research; and a case study developing a governance structure for the primary care database THIN. (slides)

Chair: Simon Ellwood-Thompson;
Panelists: David Milward (Linguamatics) – slides, Philipp Daumke (Averbis) – slides, Alison O’Neil (Canon Medical), Szymon Wartak (Babylon Health), Philip John Gorinski (Huawei Noah’s Ark lab).

Chair: Sumithra Velupillai

1. Denis R. Newman-Griffis: Kickstarting NLP for whole-person function information with representation learning and data analysis (slides)
2. Daphné A. Chopard, Matthias Treder, Irena Spasic: Text self-normalisation: An adaptive and self-sufficient approach to abbreviation expansion (slides)
3. Natalia Viani: Exploring alternative approaches to address temporal information extraction for clinical use-cases (slides)

Panelists: Beatrice Alex (University of Edinburgh), Hongfang Liu (Mayo Clinic), Anoop Shah (University College London)

Gareth Morlais, Dawn Knight, Laura Arman, Irena Spasić

When a patient and the health professional who cares for them both have first languages other than English, what’s the best way to conduct consultations, keep and share records? As Healtac2019 is in the capital city of Wales, this panel looks at the Welsh language. ‘Mwy na geiriau/More than just words: A strategic framework for promoting the Welsh language in health and social care’ was published in 2012 by the Deputy Minister for Social Services. The aim of the framework was for organisations to recognise the role of language in care and that people who need services in Welsh are actively offered the option. The follow-on strategic framework 2016-2019 aim was to maintain momentum including practical support for implementing an ‘Active Offer’ in care organisation. An ‘Active Offer’ means providing a service in Welsh without someone having to ask for it. At this panel, we will be discussing technical means that can facilitate service provision in Welsh (e.g. terminological resources, software localisation, machine translation, etc.) and a variety of application areas (e.g. prescriptions, mental health, allergies, etc.).

Chair: Goran Nenadic

1. E. Ford, L. Hassan, M. Oswald. Citizens’ Jury: What access should researchers have to free-text data in health records? (slides)
2. J. Ive, S. Velupillai, N. Viani, A. Roberts, R. Stewart, S. Puntis, W.O. Pickrell, R.N. Cardinal. Towards shareable data in clinical NLP: Generating synthetic electronic health records (slides)
3. A. Belz, R. Hoile, A. Mullick, E. Ford, J. Cassell, H. van Marwijk, D. Weir. Conceptualising and quantifying the social media signal relating to non-adherence in the treatment of depression: Dataset and annotation scheme (slides)
4. W. Dixon, G. Nenadic, A. Bulcock, M. Evans, A. Anand, M. Belousov, G. Demetriou. Feasibility of text-mining to support nudging of real-time side effect reporting to drug regulators within the online health social network ‘HealthUnlocked’ (slides)

With the increase in digitalization and advancement of high throughput technologies, the life sciences, biomedicine, and healthcare are increasingly turning into data-intensive sciences. One of the greatest challenges today in today’s science is how to manage, integrate, and mine biomedical data today resulting from various sources in different structural dimensions, ranging from the microscopic world (e.g., -omics data) to the macroscopic world (e.g., population health). Systematic and comprehensive exploration of all these data provides a mechanism for data-driven hypothesis generation, optimized experiment planning, precision medicine, and evidence-based healthcare delivery. In this talk, I will discuss our efforts towards digital health sciences illustrated through the discovery, translation, and application (DTA) paradigm. (slides)

Yunfei Long, Elvira Perez Vallejos, Mat Rawsthorne, Angus Roberts, Harish Tayyar Madabushi, Maria Liakata, Rina Dutta

This panel will have four presentations and time for questions at the end. Discussion will focus on current progress, potential and pitfalls of using natural language processing (NLP) technologies in mental health, for both clinical and non-clinical data. Special consideration will be given to further understanding how NLP related predictive and generative applications will impact on mental health service. Attention will also be placed on the ethical implications of extracting information from mental health service data and service involvement to design automated or semi-automated techniques to moderate discussions and the technical challenges d limited access and heterogeneous types of data. We would also like to discuss the connection of mental health NLP and other aspects of NLP applications. Finally, possible avenues of research are discussed. We have also envisioned on collaborating with industries to expand accessible data, enhance the computation power, and make researches more applicable to the whole population. Slides: Liakata, Rawsthorne, Dutta, XenZone.